inode.c 150.0 KB
Newer Older
1
/*
2
 *  linux/fs/ext4/inode.c
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
 *
 * Copyright (C) 1992, 1993, 1994, 1995
 * Remy Card (card@masi.ibp.fr)
 * Laboratoire MASI - Institut Blaise Pascal
 * Universite Pierre et Marie Curie (Paris VI)
 *
 *  from
 *
 *  linux/fs/minix/inode.c
 *
 *  Copyright (C) 1991, 1992  Linus Torvalds
 *
 *  64-bit file support on 64-bit platforms by Jakub Jelinek
 *	(jj@sunsite.ms.mff.cuni.cz)
 *
18
 *  Assorted race fixes, rewrite of ext4_get_block() by Al Viro, 2000
19 20 21 22
 */

#include <linux/fs.h>
#include <linux/time.h>
23
#include <linux/jbd2.h>
24 25 26 27 28 29
#include <linux/highuid.h>
#include <linux/pagemap.h>
#include <linux/quotaops.h>
#include <linux/string.h>
#include <linux/buffer_head.h>
#include <linux/writeback.h>
30
#include <linux/pagevec.h>
31
#include <linux/mpage.h>
32
#include <linux/namei.h>
33 34
#include <linux/uio.h>
#include <linux/bio.h>
35
#include <linux/workqueue.h>
36
#include <linux/kernel.h>
37
#include <linux/printk.h>
38
#include <linux/slab.h>
39
#include <linux/ratelimit.h>
40
#include <linux/aio.h>
41

42
#include "ext4_jbd2.h"
43 44
#include "xattr.h"
#include "acl.h"
45
#include "truncate.h"
46

47 48
#include <trace/events/ext4.h>

49 50
#define MPAGE_DA_EXTENT_TAIL 0x01

51 52 53 54 55 56 57 58
static __u32 ext4_inode_csum(struct inode *inode, struct ext4_inode *raw,
			      struct ext4_inode_info *ei)
{
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
	__u16 csum_lo;
	__u16 csum_hi = 0;
	__u32 csum;

59
	csum_lo = le16_to_cpu(raw->i_checksum_lo);
60 61 62
	raw->i_checksum_lo = 0;
	if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE &&
	    EXT4_FITS_IN_INODE(raw, ei, i_checksum_hi)) {
63
		csum_hi = le16_to_cpu(raw->i_checksum_hi);
64 65 66 67 68 69
		raw->i_checksum_hi = 0;
	}

	csum = ext4_chksum(sbi, ei->i_csum_seed, (__u8 *)raw,
			   EXT4_INODE_SIZE(inode->i_sb));

70
	raw->i_checksum_lo = cpu_to_le16(csum_lo);
71 72
	if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE &&
	    EXT4_FITS_IN_INODE(raw, ei, i_checksum_hi))
73
		raw->i_checksum_hi = cpu_to_le16(csum_hi);
74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117

	return csum;
}

static int ext4_inode_csum_verify(struct inode *inode, struct ext4_inode *raw,
				  struct ext4_inode_info *ei)
{
	__u32 provided, calculated;

	if (EXT4_SB(inode->i_sb)->s_es->s_creator_os !=
	    cpu_to_le32(EXT4_OS_LINUX) ||
	    !EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
		EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
		return 1;

	provided = le16_to_cpu(raw->i_checksum_lo);
	calculated = ext4_inode_csum(inode, raw, ei);
	if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE &&
	    EXT4_FITS_IN_INODE(raw, ei, i_checksum_hi))
		provided |= ((__u32)le16_to_cpu(raw->i_checksum_hi)) << 16;
	else
		calculated &= 0xFFFF;

	return provided == calculated;
}

static void ext4_inode_csum_set(struct inode *inode, struct ext4_inode *raw,
				struct ext4_inode_info *ei)
{
	__u32 csum;

	if (EXT4_SB(inode->i_sb)->s_es->s_creator_os !=
	    cpu_to_le32(EXT4_OS_LINUX) ||
	    !EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
		EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
		return;

	csum = ext4_inode_csum(inode, raw, ei);
	raw->i_checksum_lo = cpu_to_le16(csum & 0xFFFF);
	if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE &&
	    EXT4_FITS_IN_INODE(raw, ei, i_checksum_hi))
		raw->i_checksum_hi = cpu_to_le16(csum >> 16);
}

118 119 120
static inline int ext4_begin_ordered_truncate(struct inode *inode,
					      loff_t new_size)
{
121
	trace_ext4_begin_ordered_truncate(inode, new_size);
122 123 124 125 126 127 128 129 130 131 132
	/*
	 * If jinode is zero, then we never opened the file for
	 * writing, so there's no need to call
	 * jbd2_journal_begin_ordered_truncate() since there's no
	 * outstanding writes we need to flush.
	 */
	if (!EXT4_I(inode)->jinode)
		return 0;
	return jbd2_journal_begin_ordered_truncate(EXT4_JOURNAL(inode),
						   EXT4_I(inode)->jinode,
						   new_size);
133 134
}

135 136
static void ext4_invalidatepage(struct page *page, unsigned int offset,
				unsigned int length);
137 138
static int __ext4_journalled_writepage(struct page *page, unsigned int len);
static int ext4_bh_delay_or_unwritten(handle_t *handle, struct buffer_head *bh);
139 140
static int ext4_meta_trans_blocks(struct inode *inode, int lblocks,
				  int pextents);
141

142 143 144
/*
 * Test whether an inode is a fast symlink.
 */
145
static int ext4_inode_is_fast_symlink(struct inode *inode)
146
{
147
	int ea_blocks = EXT4_I(inode)->i_file_acl ?
148 149 150 151 152 153 154 155 156 157
		(inode->i_sb->s_blocksize >> 9) : 0;

	return (S_ISLNK(inode->i_mode) && inode->i_blocks - ea_blocks == 0);
}

/*
 * Restart the transaction associated with *handle.  This does a commit,
 * so before we call here everything must be consistently dirtied against
 * this transaction.
 */
158
int ext4_truncate_restart_trans(handle_t *handle, struct inode *inode,
159
				 int nblocks)
160
{
161 162 163
	int ret;

	/*
164
	 * Drop i_data_sem to avoid deadlock with ext4_map_blocks.  At this
165 166 167 168
	 * moment, get_block can be called only for blocks inside i_size since
	 * page cache has been already dropped and writes are blocked by
	 * i_mutex. So we can safely drop the i_data_sem here.
	 */
169
	BUG_ON(EXT4_JOURNAL(inode) == NULL);
170
	jbd_debug(2, "restarting handle %p\n", handle);
171
	up_write(&EXT4_I(inode)->i_data_sem);
172
	ret = ext4_journal_restart(handle, nblocks);
173
	down_write(&EXT4_I(inode)->i_data_sem);
174
	ext4_discard_preallocations(inode);
175 176

	return ret;
177 178 179 180 181
}

/*
 * Called at the last iput() if i_nlink is zero.
 */
A
Al Viro 已提交
182
void ext4_evict_inode(struct inode *inode)
183 184
{
	handle_t *handle;
185
	int err;
186

187
	trace_ext4_evict_inode(inode);
188

A
Al Viro 已提交
189
	if (inode->i_nlink) {
190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208
		/*
		 * When journalling data dirty buffers are tracked only in the
		 * journal. So although mm thinks everything is clean and
		 * ready for reaping the inode might still have some pages to
		 * write in the running transaction or waiting to be
		 * checkpointed. Thus calling jbd2_journal_invalidatepage()
		 * (via truncate_inode_pages()) to discard these buffers can
		 * cause data loss. Also even if we did not discard these
		 * buffers, we would have no way to find them after the inode
		 * is reaped and thus user could see stale data if he tries to
		 * read them before the transaction is checkpointed. So be
		 * careful and force everything to disk here... We use
		 * ei->i_datasync_tid to store the newest transaction
		 * containing inode's data.
		 *
		 * Note that directories do not have this problem because they
		 * don't use page cache.
		 */
		if (ext4_should_journal_data(inode) &&
209 210
		    (S_ISLNK(inode->i_mode) || S_ISREG(inode->i_mode)) &&
		    inode->i_ino != EXT4_JOURNAL_INO) {
211 212 213
			journal_t *journal = EXT4_SB(inode->i_sb)->s_journal;
			tid_t commit_tid = EXT4_I(inode)->i_datasync_tid;

214
			jbd2_complete_transaction(journal, commit_tid);
215 216
			filemap_write_and_wait(&inode->i_data);
		}
A
Al Viro 已提交
217
		truncate_inode_pages(&inode->i_data, 0);
J
Jan Kara 已提交
218 219

		WARN_ON(atomic_read(&EXT4_I(inode)->i_ioend_count));
A
Al Viro 已提交
220 221 222
		goto no_delete;
	}

223
	if (!is_bad_inode(inode))
224
		dquot_initialize(inode);
225

226 227
	if (ext4_should_order_data(inode))
		ext4_begin_ordered_truncate(inode, 0);
228 229
	truncate_inode_pages(&inode->i_data, 0);

J
Jan Kara 已提交
230
	WARN_ON(atomic_read(&EXT4_I(inode)->i_ioend_count));
231 232 233
	if (is_bad_inode(inode))
		goto no_delete;

234 235 236 237 238
	/*
	 * Protect us against freezing - iput() caller didn't have to have any
	 * protection against it
	 */
	sb_start_intwrite(inode->i_sb);
239 240
	handle = ext4_journal_start(inode, EXT4_HT_TRUNCATE,
				    ext4_blocks_for_truncate(inode)+3);
241
	if (IS_ERR(handle)) {
242
		ext4_std_error(inode->i_sb, PTR_ERR(handle));
243 244 245 246 247
		/*
		 * If we're going to skip the normal cleanup, we still need to
		 * make sure that the in-core orphan linked list is properly
		 * cleaned up.
		 */
248
		ext4_orphan_del(NULL, inode);
249
		sb_end_intwrite(inode->i_sb);
250 251 252 253
		goto no_delete;
	}

	if (IS_SYNC(inode))
254
		ext4_handle_sync(handle);
255
	inode->i_size = 0;
256 257
	err = ext4_mark_inode_dirty(handle, inode);
	if (err) {
258
		ext4_warning(inode->i_sb,
259 260 261
			     "couldn't mark inode dirty (err %d)", err);
		goto stop_handle;
	}
262
	if (inode->i_blocks)
263
		ext4_truncate(inode);
264 265 266 267 268 269 270

	/*
	 * ext4_ext_truncate() doesn't reserve any slop when it
	 * restarts journal transactions; therefore there may not be
	 * enough credits left in the handle to remove the inode from
	 * the orphan list and set the dtime field.
	 */
271
	if (!ext4_handle_has_enough_credits(handle, 3)) {
272 273 274 275
		err = ext4_journal_extend(handle, 3);
		if (err > 0)
			err = ext4_journal_restart(handle, 3);
		if (err != 0) {
276
			ext4_warning(inode->i_sb,
277 278 279
				     "couldn't extend journal (err %d)", err);
		stop_handle:
			ext4_journal_stop(handle);
280
			ext4_orphan_del(NULL, inode);
281
			sb_end_intwrite(inode->i_sb);
282 283 284 285
			goto no_delete;
		}
	}

286
	/*
287
	 * Kill off the orphan record which ext4_truncate created.
288
	 * AKPM: I think this can be inside the above `if'.
289
	 * Note that ext4_orphan_del() has to be able to cope with the
290
	 * deletion of a non-existent orphan - this is because we don't
291
	 * know if ext4_truncate() actually created an orphan record.
292 293
	 * (Well, we could do this if we need to, but heck - it works)
	 */
294 295
	ext4_orphan_del(handle, inode);
	EXT4_I(inode)->i_dtime	= get_seconds();
296 297 298 299 300 301 302 303

	/*
	 * One subtle ordering requirement: if anything has gone wrong
	 * (transaction abort, IO errors, whatever), then we can still
	 * do these next steps (the fs will already have been marked as
	 * having errors), but we can't free the inode if the mark_dirty
	 * fails.
	 */
304
	if (ext4_mark_inode_dirty(handle, inode))
305
		/* If that failed, just do the required in-core inode clear. */
A
Al Viro 已提交
306
		ext4_clear_inode(inode);
307
	else
308 309
		ext4_free_inode(handle, inode);
	ext4_journal_stop(handle);
310
	sb_end_intwrite(inode->i_sb);
311 312
	return;
no_delete:
A
Al Viro 已提交
313
	ext4_clear_inode(inode);	/* We must guarantee clearing of inode... */
314 315
}

316 317
#ifdef CONFIG_QUOTA
qsize_t *ext4_get_reserved_space(struct inode *inode)
318
{
319
	return &EXT4_I(inode)->i_reserved_quota;
320
}
321
#endif
322

323 324
/*
 * Calculate the number of metadata blocks need to reserve
325
 * to allocate a block located at @lblock
326
 */
327
static int ext4_calc_metadata_amount(struct inode *inode, ext4_lblk_t lblock)
328
{
329
	if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
330
		return ext4_ext_calc_metadata_amount(inode, lblock);
331

332
	return ext4_ind_calc_metadata_amount(inode, lblock);
333 334
}

335 336 337 338
/*
 * Called with i_data_sem down, which is important since we can call
 * ext4_discard_preallocations() from here.
 */
339 340
void ext4_da_update_reserve_space(struct inode *inode,
					int used, int quota_claim)
341 342
{
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
343 344 345
	struct ext4_inode_info *ei = EXT4_I(inode);

	spin_lock(&ei->i_block_reservation_lock);
346
	trace_ext4_da_update_reserve_space(inode, used, quota_claim);
347
	if (unlikely(used > ei->i_reserved_data_blocks)) {
348
		ext4_warning(inode->i_sb, "%s: ino %lu, used %d "
349
			 "with only %d reserved data blocks",
350 351 352 353 354
			 __func__, inode->i_ino, used,
			 ei->i_reserved_data_blocks);
		WARN_ON(1);
		used = ei->i_reserved_data_blocks;
	}
355

356
	if (unlikely(ei->i_allocated_meta_blocks > ei->i_reserved_meta_blocks)) {
357 358 359 360 361 362
		ext4_warning(inode->i_sb, "ino %lu, allocated %d "
			"with only %d reserved metadata blocks "
			"(releasing %d blocks with reserved %d data blocks)",
			inode->i_ino, ei->i_allocated_meta_blocks,
			     ei->i_reserved_meta_blocks, used,
			     ei->i_reserved_data_blocks);
363 364 365 366
		WARN_ON(1);
		ei->i_allocated_meta_blocks = ei->i_reserved_meta_blocks;
	}

367 368 369
	/* Update per-inode reservations */
	ei->i_reserved_data_blocks -= used;
	ei->i_reserved_meta_blocks -= ei->i_allocated_meta_blocks;
370
	percpu_counter_sub(&sbi->s_dirtyclusters_counter,
371
			   used + ei->i_allocated_meta_blocks);
372
	ei->i_allocated_meta_blocks = 0;
373

374 375 376 377 378 379
	if (ei->i_reserved_data_blocks == 0) {
		/*
		 * We can release all of the reserved metadata blocks
		 * only when we have written all of the delayed
		 * allocation blocks.
		 */
380
		percpu_counter_sub(&sbi->s_dirtyclusters_counter,
381
				   ei->i_reserved_meta_blocks);
382
		ei->i_reserved_meta_blocks = 0;
383
		ei->i_da_metadata_calc_len = 0;
384
	}
385
	spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
386

387 388
	/* Update quota subsystem for data blocks */
	if (quota_claim)
389
		dquot_claim_block(inode, EXT4_C2B(sbi, used));
390
	else {
391 392 393
		/*
		 * We did fallocate with an offset that is already delayed
		 * allocated. So on delayed allocated writeback we should
394
		 * not re-claim the quota for fallocated blocks.
395
		 */
396
		dquot_release_reservation_block(inode, EXT4_C2B(sbi, used));
397
	}
398 399 400 401 402 403

	/*
	 * If we have done all the pending block allocations and if
	 * there aren't any writers on the inode, we can discard the
	 * inode's preallocations.
	 */
404 405
	if ((ei->i_reserved_data_blocks == 0) &&
	    (atomic_read(&inode->i_writecount) == 0))
406
		ext4_discard_preallocations(inode);
407 408
}

409
static int __check_block_validity(struct inode *inode, const char *func,
410 411
				unsigned int line,
				struct ext4_map_blocks *map)
412
{
413 414
	if (!ext4_data_block_valid(EXT4_SB(inode->i_sb), map->m_pblk,
				   map->m_len)) {
415 416 417 418
		ext4_error_inode(inode, func, line, map->m_pblk,
				 "lblock %lu mapped to illegal pblock "
				 "(length %d)", (unsigned long) map->m_lblk,
				 map->m_len);
419 420 421 422 423
		return -EIO;
	}
	return 0;
}

424
#define check_block_validity(inode, map)	\
425
	__check_block_validity((inode), __func__, __LINE__, (map))
426

427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467
#ifdef ES_AGGRESSIVE_TEST
static void ext4_map_blocks_es_recheck(handle_t *handle,
				       struct inode *inode,
				       struct ext4_map_blocks *es_map,
				       struct ext4_map_blocks *map,
				       int flags)
{
	int retval;

	map->m_flags = 0;
	/*
	 * There is a race window that the result is not the same.
	 * e.g. xfstests #223 when dioread_nolock enables.  The reason
	 * is that we lookup a block mapping in extent status tree with
	 * out taking i_data_sem.  So at the time the unwritten extent
	 * could be converted.
	 */
	if (!(flags & EXT4_GET_BLOCKS_NO_LOCK))
		down_read((&EXT4_I(inode)->i_data_sem));
	if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
		retval = ext4_ext_map_blocks(handle, inode, map, flags &
					     EXT4_GET_BLOCKS_KEEP_SIZE);
	} else {
		retval = ext4_ind_map_blocks(handle, inode, map, flags &
					     EXT4_GET_BLOCKS_KEEP_SIZE);
	}
	if (!(flags & EXT4_GET_BLOCKS_NO_LOCK))
		up_read((&EXT4_I(inode)->i_data_sem));
	/*
	 * Clear EXT4_MAP_FROM_CLUSTER and EXT4_MAP_BOUNDARY flag
	 * because it shouldn't be marked in es_map->m_flags.
	 */
	map->m_flags &= ~(EXT4_MAP_FROM_CLUSTER | EXT4_MAP_BOUNDARY);

	/*
	 * We don't check m_len because extent will be collpased in status
	 * tree.  So the m_len might not equal.
	 */
	if (es_map->m_lblk != map->m_lblk ||
	    es_map->m_flags != map->m_flags ||
	    es_map->m_pblk != map->m_pblk) {
468
		printk("ES cache assertion failed for inode: %lu "
469 470 471 472 473 474 475 476 477 478
		       "es_cached ex [%d/%d/%llu/%x] != "
		       "found ex [%d/%d/%llu/%x] retval %d flags %x\n",
		       inode->i_ino, es_map->m_lblk, es_map->m_len,
		       es_map->m_pblk, es_map->m_flags, map->m_lblk,
		       map->m_len, map->m_pblk, map->m_flags,
		       retval, flags);
	}
}
#endif /* ES_AGGRESSIVE_TEST */

479
/*
480
 * The ext4_map_blocks() function tries to look up the requested blocks,
481
 * and returns if the blocks are already mapped.
482 483 484 485 486
 *
 * Otherwise it takes the write lock of the i_data_sem and allocate blocks
 * and store the allocated blocks in the result buffer head and mark it
 * mapped.
 *
487 488
 * If file type is extents based, it will call ext4_ext_map_blocks(),
 * Otherwise, call with ext4_ind_map_blocks() to handle indirect mapping
489 490 491 492 493 494 495 496
 * based files
 *
 * On success, it returns the number of blocks being mapped or allocate.
 * if create==0 and the blocks are pre-allocated and uninitialized block,
 * the result buffer head is unmapped. If the create ==1, it will make sure
 * the buffer head is mapped.
 *
 * It returns 0 if plain look up failed (blocks have not been allocated), in
497
 * that case, buffer head is unmapped
498 499 500
 *
 * It returns the error in case of allocation failure.
 */
501 502
int ext4_map_blocks(handle_t *handle, struct inode *inode,
		    struct ext4_map_blocks *map, int flags)
503
{
504
	struct extent_status es;
505
	int retval;
506 507 508 509 510
#ifdef ES_AGGRESSIVE_TEST
	struct ext4_map_blocks orig_map;

	memcpy(&orig_map, map, sizeof(*map));
#endif
511

512 513 514 515
	map->m_flags = 0;
	ext_debug("ext4_map_blocks(): inode %lu, flag %d, max_blocks %u,"
		  "logical block %lu\n", inode->i_ino, flags, map->m_len,
		  (unsigned long) map->m_lblk);
516 517 518

	/* Lookup extent status tree firstly */
	if (ext4_es_lookup_extent(inode, map->m_lblk, &es)) {
519
		ext4_es_lru_add(inode);
520 521 522 523 524 525 526 527 528 529 530 531 532 533
		if (ext4_es_is_written(&es) || ext4_es_is_unwritten(&es)) {
			map->m_pblk = ext4_es_pblock(&es) +
					map->m_lblk - es.es_lblk;
			map->m_flags |= ext4_es_is_written(&es) ?
					EXT4_MAP_MAPPED : EXT4_MAP_UNWRITTEN;
			retval = es.es_len - (map->m_lblk - es.es_lblk);
			if (retval > map->m_len)
				retval = map->m_len;
			map->m_len = retval;
		} else if (ext4_es_is_delayed(&es) || ext4_es_is_hole(&es)) {
			retval = 0;
		} else {
			BUG_ON(1);
		}
534 535 536 537
#ifdef ES_AGGRESSIVE_TEST
		ext4_map_blocks_es_recheck(handle, inode, map,
					   &orig_map, flags);
#endif
538 539 540
		goto found;
	}

541
	/*
542 543
	 * Try to see if we can get the block without requesting a new
	 * file system block.
544
	 */
545 546
	if (!(flags & EXT4_GET_BLOCKS_NO_LOCK))
		down_read((&EXT4_I(inode)->i_data_sem));
547
	if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
548 549
		retval = ext4_ext_map_blocks(handle, inode, map, flags &
					     EXT4_GET_BLOCKS_KEEP_SIZE);
550
	} else {
551 552
		retval = ext4_ind_map_blocks(handle, inode, map, flags &
					     EXT4_GET_BLOCKS_KEEP_SIZE);
553
	}
554 555
	if (retval > 0) {
		int ret;
556
		unsigned int status;
557

558 559 560 561 562 563
		if (unlikely(retval != map->m_len)) {
			ext4_warning(inode->i_sb,
				     "ES len assertion failed for inode "
				     "%lu: retval %d != map->m_len %d",
				     inode->i_ino, retval, map->m_len);
			WARN_ON(1);
564 565
		}

566 567 568 569 570 571 572 573 574 575 576
		status = map->m_flags & EXT4_MAP_UNWRITTEN ?
				EXTENT_STATUS_UNWRITTEN : EXTENT_STATUS_WRITTEN;
		if (!(flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE) &&
		    ext4_find_delalloc_range(inode, map->m_lblk,
					     map->m_lblk + map->m_len - 1))
			status |= EXTENT_STATUS_DELAYED;
		ret = ext4_es_insert_extent(inode, map->m_lblk,
					    map->m_len, map->m_pblk, status);
		if (ret < 0)
			retval = ret;
	}
577 578
	if (!(flags & EXT4_GET_BLOCKS_NO_LOCK))
		up_read((&EXT4_I(inode)->i_data_sem));
579

580
found:
581
	if (retval > 0 && map->m_flags & EXT4_MAP_MAPPED) {
582
		int ret = check_block_validity(inode, map);
583 584 585 586
		if (ret != 0)
			return ret;
	}

587
	/* If it is only a block(s) look up */
588
	if ((flags & EXT4_GET_BLOCKS_CREATE) == 0)
589 590 591 592 593 594
		return retval;

	/*
	 * Returns if the blocks have already allocated
	 *
	 * Note that if blocks have been preallocated
595
	 * ext4_ext_get_block() returns the create = 0
596 597
	 * with buffer head unmapped.
	 */
598
	if (retval > 0 && map->m_flags & EXT4_MAP_MAPPED)
599 600
		return retval;

601
	/*
602 603
	 * Here we clear m_flags because after allocating an new extent,
	 * it will be set again.
604
	 */
605
	map->m_flags &= ~EXT4_MAP_FLAGS;
606

607
	/*
608 609 610 611
	 * New blocks allocate and/or writing to uninitialized extent
	 * will possibly result in updating i_data, so we take
	 * the write lock of i_data_sem, and call get_blocks()
	 * with create == 1 flag.
612 613
	 */
	down_write((&EXT4_I(inode)->i_data_sem));
614 615 616 617 618 619 620

	/*
	 * if the caller is from delayed allocation writeout path
	 * we have already reserved fs blocks for allocation
	 * let the underlying get_block() function know to
	 * avoid double accounting
	 */
621
	if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE)
622
		ext4_set_inode_state(inode, EXT4_STATE_DELALLOC_RESERVED);
623 624 625 626
	/*
	 * We need to check for EXT4 here because migrate
	 * could have changed the inode type in between
	 */
627
	if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
628
		retval = ext4_ext_map_blocks(handle, inode, map, flags);
629
	} else {
630
		retval = ext4_ind_map_blocks(handle, inode, map, flags);
631

632
		if (retval > 0 && map->m_flags & EXT4_MAP_NEW) {
633 634 635 636 637
			/*
			 * We allocated new blocks which will result in
			 * i_data's format changing.  Force the migrate
			 * to fail by clearing migrate flags
			 */
638
			ext4_clear_inode_state(inode, EXT4_STATE_EXT_MIGRATE);
639
		}
640

641 642 643 644 645 646 647
		/*
		 * Update reserved blocks/metadata blocks after successful
		 * block allocation which had been deferred till now. We don't
		 * support fallocate for non extent files. So we can update
		 * reserve space here.
		 */
		if ((retval > 0) &&
648
			(flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE))
649 650
			ext4_da_update_reserve_space(inode, retval, 1);
	}
651
	if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE)
652
		ext4_clear_inode_state(inode, EXT4_STATE_DELALLOC_RESERVED);
653

654 655
	if (retval > 0) {
		int ret;
656
		unsigned int status;
657

658 659 660 661 662 663
		if (unlikely(retval != map->m_len)) {
			ext4_warning(inode->i_sb,
				     "ES len assertion failed for inode "
				     "%lu: retval %d != map->m_len %d",
				     inode->i_ino, retval, map->m_len);
			WARN_ON(1);
664 665
		}

666 667 668 669 670 671 672 673 674
		/*
		 * If the extent has been zeroed out, we don't need to update
		 * extent status tree.
		 */
		if ((flags & EXT4_GET_BLOCKS_PRE_IO) &&
		    ext4_es_lookup_extent(inode, map->m_lblk, &es)) {
			if (ext4_es_is_written(&es))
				goto has_zeroout;
		}
675 676 677 678 679 680 681 682 683 684
		status = map->m_flags & EXT4_MAP_UNWRITTEN ?
				EXTENT_STATUS_UNWRITTEN : EXTENT_STATUS_WRITTEN;
		if (!(flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE) &&
		    ext4_find_delalloc_range(inode, map->m_lblk,
					     map->m_lblk + map->m_len - 1))
			status |= EXTENT_STATUS_DELAYED;
		ret = ext4_es_insert_extent(inode, map->m_lblk, map->m_len,
					    map->m_pblk, status);
		if (ret < 0)
			retval = ret;
685 686
	}

687
has_zeroout:
688
	up_write((&EXT4_I(inode)->i_data_sem));
689
	if (retval > 0 && map->m_flags & EXT4_MAP_MAPPED) {
690
		int ret = check_block_validity(inode, map);
691 692 693
		if (ret != 0)
			return ret;
	}
694 695 696
	return retval;
}

697 698 699
/* Maximum number of blocks we map for direct IO at once. */
#define DIO_MAX_BLOCKS 4096

700 701
static int _ext4_get_block(struct inode *inode, sector_t iblock,
			   struct buffer_head *bh, int flags)
702
{
703
	handle_t *handle = ext4_journal_current_handle();
704
	struct ext4_map_blocks map;
J
Jan Kara 已提交
705
	int ret = 0, started = 0;
706
	int dio_credits;
707

T
Tao Ma 已提交
708 709 710
	if (ext4_has_inline_data(inode))
		return -ERANGE;

711 712 713
	map.m_lblk = iblock;
	map.m_len = bh->b_size >> inode->i_blkbits;

714
	if (flags && !(flags & EXT4_GET_BLOCKS_NO_LOCK) && !handle) {
J
Jan Kara 已提交
715
		/* Direct IO write... */
716 717 718
		if (map.m_len > DIO_MAX_BLOCKS)
			map.m_len = DIO_MAX_BLOCKS;
		dio_credits = ext4_chunk_trans_blocks(inode, map.m_len);
719 720
		handle = ext4_journal_start(inode, EXT4_HT_MAP_BLOCKS,
					    dio_credits);
J
Jan Kara 已提交
721
		if (IS_ERR(handle)) {
722
			ret = PTR_ERR(handle);
723
			return ret;
724
		}
J
Jan Kara 已提交
725
		started = 1;
726 727
	}

728
	ret = ext4_map_blocks(handle, inode, &map, flags);
J
Jan Kara 已提交
729
	if (ret > 0) {
730 731 732
		map_bh(bh, inode->i_sb, map.m_pblk);
		bh->b_state = (bh->b_state & ~EXT4_MAP_FLAGS) | map.m_flags;
		bh->b_size = inode->i_sb->s_blocksize * map.m_len;
J
Jan Kara 已提交
733
		ret = 0;
734
	}
J
Jan Kara 已提交
735 736
	if (started)
		ext4_journal_stop(handle);
737 738 739
	return ret;
}

740 741 742 743 744 745 746
int ext4_get_block(struct inode *inode, sector_t iblock,
		   struct buffer_head *bh, int create)
{
	return _ext4_get_block(inode, iblock, bh,
			       create ? EXT4_GET_BLOCKS_CREATE : 0);
}

747 748 749
/*
 * `handle' can be NULL if create is zero
 */
750
struct buffer_head *ext4_getblk(handle_t *handle, struct inode *inode,
A
Aneesh Kumar K.V 已提交
751
				ext4_lblk_t block, int create, int *errp)
752
{
753 754
	struct ext4_map_blocks map;
	struct buffer_head *bh;
755 756 757 758
	int fatal = 0, err;

	J_ASSERT(handle != NULL || create == 0);

759 760 761 762
	map.m_lblk = block;
	map.m_len = 1;
	err = ext4_map_blocks(handle, inode, &map,
			      create ? EXT4_GET_BLOCKS_CREATE : 0);
763

764 765 766
	/* ensure we send some value back into *errp */
	*errp = 0;

767 768
	if (create && err == 0)
		err = -ENOSPC;	/* should never happen */
769 770 771 772 773 774
	if (err < 0)
		*errp = err;
	if (err <= 0)
		return NULL;

	bh = sb_getblk(inode->i_sb, map.m_pblk);
775
	if (unlikely(!bh)) {
776
		*errp = -ENOMEM;
777
		return NULL;
778
	}
779 780 781
	if (map.m_flags & EXT4_MAP_NEW) {
		J_ASSERT(create != 0);
		J_ASSERT(handle != NULL);
782

783 784 785 786 787 788 789 790 791 792 793 794 795
		/*
		 * Now that we do not always journal data, we should
		 * keep in mind whether this should always journal the
		 * new buffer as metadata.  For now, regular file
		 * writes use ext4_get_block instead, so it's not a
		 * problem.
		 */
		lock_buffer(bh);
		BUFFER_TRACE(bh, "call get_create_access");
		fatal = ext4_journal_get_create_access(handle, bh);
		if (!fatal && !buffer_uptodate(bh)) {
			memset(bh->b_data, 0, inode->i_sb->s_blocksize);
			set_buffer_uptodate(bh);
796
		}
797 798 799 800 801 802 803
		unlock_buffer(bh);
		BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
		err = ext4_handle_dirty_metadata(handle, inode, bh);
		if (!fatal)
			fatal = err;
	} else {
		BUFFER_TRACE(bh, "not a new buffer");
804
	}
805 806 807 808 809 810
	if (fatal) {
		*errp = fatal;
		brelse(bh);
		bh = NULL;
	}
	return bh;
811 812
}

813
struct buffer_head *ext4_bread(handle_t *handle, struct inode *inode,
A
Aneesh Kumar K.V 已提交
814
			       ext4_lblk_t block, int create, int *err)
815
{
816
	struct buffer_head *bh;
817

818
	bh = ext4_getblk(handle, inode, block, create, err);
819 820 821 822
	if (!bh)
		return bh;
	if (buffer_uptodate(bh))
		return bh;
823
	ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &bh);
824 825 826 827 828 829 830 831
	wait_on_buffer(bh);
	if (buffer_uptodate(bh))
		return bh;
	put_bh(bh);
	*err = -EIO;
	return NULL;
}

832 833 834 835 836 837 838
int ext4_walk_page_buffers(handle_t *handle,
			   struct buffer_head *head,
			   unsigned from,
			   unsigned to,
			   int *partial,
			   int (*fn)(handle_t *handle,
				     struct buffer_head *bh))
839 840 841 842 843 844 845
{
	struct buffer_head *bh;
	unsigned block_start, block_end;
	unsigned blocksize = head->b_size;
	int err, ret = 0;
	struct buffer_head *next;

846 847
	for (bh = head, block_start = 0;
	     ret == 0 && (bh != head || !block_start);
848
	     block_start = block_end, bh = next) {
849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865
		next = bh->b_this_page;
		block_end = block_start + blocksize;
		if (block_end <= from || block_start >= to) {
			if (partial && !buffer_uptodate(bh))
				*partial = 1;
			continue;
		}
		err = (*fn)(handle, bh);
		if (!ret)
			ret = err;
	}
	return ret;
}

/*
 * To preserve ordering, it is essential that the hole instantiation and
 * the data write be encapsulated in a single transaction.  We cannot
866
 * close off a transaction and start a new one between the ext4_get_block()
867
 * and the commit_write().  So doing the jbd2_journal_start at the start of
868 869
 * prepare_write() is the right place.
 *
870 871 872 873
 * Also, this function can nest inside ext4_writepage().  In that case, we
 * *know* that ext4_writepage() has generated enough buffer credits to do the
 * whole page.  So we won't block on the journal in that case, which is good,
 * because the caller may be PF_MEMALLOC.
874
 *
875
 * By accident, ext4 can be reentered when a transaction is open via
876 877 878 879 880 881
 * quota file writes.  If we were to commit the transaction while thus
 * reentered, there can be a deadlock - we would be holding a quota
 * lock, and the commit would never complete if another thread had a
 * transaction open and was blocking on the quota lock - a ranking
 * violation.
 *
882
 * So what we do is to rely on the fact that jbd2_journal_stop/journal_start
883 884 885 886
 * will _not_ run commit under these circumstances because handle->h_ref
 * is elevated.  We'll still have enough credits for the tiny quotafile
 * write.
 */
887 888
int do_journal_get_write_access(handle_t *handle,
				struct buffer_head *bh)
889
{
890 891 892
	int dirty = buffer_dirty(bh);
	int ret;

893 894
	if (!buffer_mapped(bh) || buffer_freed(bh))
		return 0;
895
	/*
C
Christoph Hellwig 已提交
896
	 * __block_write_begin() could have dirtied some buffers. Clean
897 898
	 * the dirty bit as jbd2_journal_get_write_access() could complain
	 * otherwise about fs integrity issues. Setting of the dirty bit
C
Christoph Hellwig 已提交
899
	 * by __block_write_begin() isn't a real problem here as we clear
900 901 902 903 904 905 906 907 908
	 * the bit before releasing a page lock and thus writeback cannot
	 * ever write the buffer.
	 */
	if (dirty)
		clear_buffer_dirty(bh);
	ret = ext4_journal_get_write_access(handle, bh);
	if (!ret && dirty)
		ret = ext4_handle_dirty_metadata(handle, NULL, bh);
	return ret;
909 910
}

911 912
static int ext4_get_block_write_nolock(struct inode *inode, sector_t iblock,
		   struct buffer_head *bh_result, int create);
N
Nick Piggin 已提交
913
static int ext4_write_begin(struct file *file, struct address_space *mapping,
914 915
			    loff_t pos, unsigned len, unsigned flags,
			    struct page **pagep, void **fsdata)
916
{
917
	struct inode *inode = mapping->host;
918
	int ret, needed_blocks;
919 920
	handle_t *handle;
	int retries = 0;
921
	struct page *page;
922
	pgoff_t index;
923
	unsigned from, to;
N
Nick Piggin 已提交
924

925
	trace_ext4_write_begin(inode, pos, len, flags);
926 927 928 929 930
	/*
	 * Reserve one block more for addition to orphan list in case
	 * we allocate blocks but write fails for some reason
	 */
	needed_blocks = ext4_writepage_trans_blocks(inode) + 1;
931
	index = pos >> PAGE_CACHE_SHIFT;
932 933
	from = pos & (PAGE_CACHE_SIZE - 1);
	to = from + len;
934

935 936 937 938
	if (ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA)) {
		ret = ext4_try_to_write_inline_data(mapping, inode, pos, len,
						    flags, pagep);
		if (ret < 0)
939 940 941
			return ret;
		if (ret == 1)
			return 0;
942 943
	}

944 945 946 947 948 949 950 951 952 953 954 955 956 957
	/*
	 * grab_cache_page_write_begin() can take a long time if the
	 * system is thrashing due to memory pressure, or if the page
	 * is being written back.  So grab it first before we start
	 * the transaction handle.  This also allows us to allocate
	 * the page (if needed) without using GFP_NOFS.
	 */
retry_grab:
	page = grab_cache_page_write_begin(mapping, index, flags);
	if (!page)
		return -ENOMEM;
	unlock_page(page);

retry_journal:
958
	handle = ext4_journal_start(inode, EXT4_HT_WRITE_PAGE, needed_blocks);
959
	if (IS_ERR(handle)) {
960 961
		page_cache_release(page);
		return PTR_ERR(handle);
962
	}
963

964 965 966 967 968
	lock_page(page);
	if (page->mapping != mapping) {
		/* The page got truncated from under us */
		unlock_page(page);
		page_cache_release(page);
969
		ext4_journal_stop(handle);
970
		goto retry_grab;
971
	}
972 973
	/* In case writeback began while the page was unlocked */
	wait_for_stable_page(page);
974

975
	if (ext4_should_dioread_nolock(inode))
976
		ret = __block_write_begin(page, pos, len, ext4_get_block_write);
977
	else
978
		ret = __block_write_begin(page, pos, len, ext4_get_block);
N
Nick Piggin 已提交
979 980

	if (!ret && ext4_should_journal_data(inode)) {
981 982 983
		ret = ext4_walk_page_buffers(handle, page_buffers(page),
					     from, to, NULL,
					     do_journal_get_write_access);
984
	}
N
Nick Piggin 已提交
985 986

	if (ret) {
987
		unlock_page(page);
988
		/*
989
		 * __block_write_begin may have instantiated a few blocks
990 991
		 * outside i_size.  Trim these off again. Don't need
		 * i_size_read because we hold i_mutex.
992 993 994
		 *
		 * Add inode to orphan list in case we crash before
		 * truncate finishes
995
		 */
996
		if (pos + len > inode->i_size && ext4_can_truncate(inode))
997 998 999 1000
			ext4_orphan_add(handle, inode);

		ext4_journal_stop(handle);
		if (pos + len > inode->i_size) {
1001
			ext4_truncate_failed_write(inode);
1002
			/*
1003
			 * If truncate failed early the inode might
1004 1005 1006 1007 1008 1009 1010
			 * still be on the orphan list; we need to
			 * make sure the inode is removed from the
			 * orphan list in that case.
			 */
			if (inode->i_nlink)
				ext4_orphan_del(NULL, inode);
		}
N
Nick Piggin 已提交
1011

1012 1013 1014 1015 1016 1017 1018
		if (ret == -ENOSPC &&
		    ext4_should_retry_alloc(inode->i_sb, &retries))
			goto retry_journal;
		page_cache_release(page);
		return ret;
	}
	*pagep = page;
1019 1020 1021
	return ret;
}

N
Nick Piggin 已提交
1022 1023
/* For write_end() in data=journal mode */
static int write_end_fn(handle_t *handle, struct buffer_head *bh)
1024
{
1025
	int ret;
1026 1027 1028
	if (!buffer_mapped(bh) || buffer_freed(bh))
		return 0;
	set_buffer_uptodate(bh);
1029 1030 1031 1032
	ret = ext4_handle_dirty_metadata(handle, NULL, bh);
	clear_buffer_meta(bh);
	clear_buffer_prio(bh);
	return ret;
1033 1034
}

1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045
/*
 * We need to pick up the new inode size which generic_commit_write gave us
 * `file' can be NULL - eg, when called from page_symlink().
 *
 * ext4 never places buffers on inode->i_mapping->private_list.  metadata
 * buffers are managed internally.
 */
static int ext4_write_end(struct file *file,
			  struct address_space *mapping,
			  loff_t pos, unsigned len, unsigned copied,
			  struct page *page, void *fsdata)
1046 1047
{
	handle_t *handle = ext4_journal_current_handle();
1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060
	struct inode *inode = mapping->host;
	int ret = 0, ret2;
	int i_size_changed = 0;

	trace_ext4_write_end(inode, pos, len, copied);
	if (ext4_test_inode_state(inode, EXT4_STATE_ORDERED_MODE)) {
		ret = ext4_jbd2_file_inode(handle, inode);
		if (ret) {
			unlock_page(page);
			page_cache_release(page);
			goto errout;
		}
	}
1061

1062 1063 1064 1065 1066 1067 1068
	if (ext4_has_inline_data(inode)) {
		ret = ext4_write_inline_data_end(inode, pos, len,
						 copied, page);
		if (ret < 0)
			goto errout;
		copied = ret;
	} else
1069 1070
		copied = block_write_end(file, mapping, pos,
					 len, copied, page, fsdata);
1071 1072 1073

	/*
	 * No need to use i_size_read() here, the i_size
1074
	 * cannot change under us because we hole i_mutex.
1075 1076 1077 1078 1079 1080 1081 1082 1083
	 *
	 * But it's important to update i_size while still holding page lock:
	 * page writeout could otherwise come in and zero beyond i_size.
	 */
	if (pos + copied > inode->i_size) {
		i_size_write(inode, pos + copied);
		i_size_changed = 1;
	}

1084
	if (pos + copied > EXT4_I(inode)->i_disksize) {
1085 1086
		/* We need to mark inode dirty even if
		 * new_i_size is less that inode->i_size
1087
		 * but greater than i_disksize. (hint delalloc)
1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103
		 */
		ext4_update_i_disksize(inode, (pos + copied));
		i_size_changed = 1;
	}
	unlock_page(page);
	page_cache_release(page);

	/*
	 * Don't mark the inode dirty under page lock. First, it unnecessarily
	 * makes the holding time of page lock longer. Second, it forces lock
	 * ordering of page lock and transaction start for journaling
	 * filesystems.
	 */
	if (i_size_changed)
		ext4_mark_inode_dirty(handle, inode);

1104
	if (pos + len > inode->i_size && ext4_can_truncate(inode))
1105 1106 1107 1108 1109
		/* if we have allocated more blocks and copied
		 * less. We will have blocks allocated outside
		 * inode->i_size. So truncate them
		 */
		ext4_orphan_add(handle, inode);
1110
errout:
1111
	ret2 = ext4_journal_stop(handle);
1112 1113
	if (!ret)
		ret = ret2;
N
Nick Piggin 已提交
1114

1115
	if (pos + len > inode->i_size) {
1116
		ext4_truncate_failed_write(inode);
1117
		/*
1118
		 * If truncate failed early the inode might still be
1119 1120 1121 1122 1123 1124 1125
		 * on the orphan list; we need to make sure the inode
		 * is removed from the orphan list in that case.
		 */
		if (inode->i_nlink)
			ext4_orphan_del(NULL, inode);
	}

N
Nick Piggin 已提交
1126
	return ret ? ret : copied;
1127 1128
}

N
Nick Piggin 已提交
1129
static int ext4_journalled_write_end(struct file *file,
1130 1131 1132
				     struct address_space *mapping,
				     loff_t pos, unsigned len, unsigned copied,
				     struct page *page, void *fsdata)
1133
{
1134
	handle_t *handle = ext4_journal_current_handle();
N
Nick Piggin 已提交
1135
	struct inode *inode = mapping->host;
1136 1137
	int ret = 0, ret2;
	int partial = 0;
N
Nick Piggin 已提交
1138
	unsigned from, to;
1139
	loff_t new_i_size;
1140

1141
	trace_ext4_journalled_write_end(inode, pos, len, copied);
N
Nick Piggin 已提交
1142 1143 1144
	from = pos & (PAGE_CACHE_SIZE - 1);
	to = from + len;

1145 1146
	BUG_ON(!ext4_handle_valid(handle));

1147 1148 1149 1150 1151 1152 1153 1154 1155
	if (ext4_has_inline_data(inode))
		copied = ext4_write_inline_data_end(inode, pos, len,
						    copied, page);
	else {
		if (copied < len) {
			if (!PageUptodate(page))
				copied = 0;
			page_zero_new_buffers(page, from+copied, to);
		}
1156

1157 1158 1159 1160 1161
		ret = ext4_walk_page_buffers(handle, page_buffers(page), from,
					     to, &partial, write_end_fn);
		if (!partial)
			SetPageUptodate(page);
	}
1162 1163
	new_i_size = pos + copied;
	if (new_i_size > inode->i_size)
N
Nick Piggin 已提交
1164
		i_size_write(inode, pos+copied);
1165
	ext4_set_inode_state(inode, EXT4_STATE_JDATA);
1166
	EXT4_I(inode)->i_datasync_tid = handle->h_transaction->t_tid;
1167 1168
	if (new_i_size > EXT4_I(inode)->i_disksize) {
		ext4_update_i_disksize(inode, new_i_size);
1169
		ret2 = ext4_mark_inode_dirty(handle, inode);
1170 1171 1172
		if (!ret)
			ret = ret2;
	}
N
Nick Piggin 已提交
1173

1174
	unlock_page(page);
1175
	page_cache_release(page);
1176
	if (pos + len > inode->i_size && ext4_can_truncate(inode))
1177 1178 1179 1180 1181 1182
		/* if we have allocated more blocks and copied
		 * less. We will have blocks allocated outside
		 * inode->i_size. So truncate them
		 */
		ext4_orphan_add(handle, inode);

1183
	ret2 = ext4_journal_stop(handle);
1184 1185
	if (!ret)
		ret = ret2;
1186
	if (pos + len > inode->i_size) {
1187
		ext4_truncate_failed_write(inode);
1188
		/*
1189
		 * If truncate failed early the inode might still be
1190 1191 1192 1193 1194 1195
		 * on the orphan list; we need to make sure the inode
		 * is removed from the orphan list in that case.
		 */
		if (inode->i_nlink)
			ext4_orphan_del(NULL, inode);
	}
N
Nick Piggin 已提交
1196 1197

	return ret ? ret : copied;
1198
}
1199

1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248
/*
 * Reserve a metadata for a single block located at lblock
 */
static int ext4_da_reserve_metadata(struct inode *inode, ext4_lblk_t lblock)
{
	int retries = 0;
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
	struct ext4_inode_info *ei = EXT4_I(inode);
	unsigned int md_needed;
	ext4_lblk_t save_last_lblock;
	int save_len;

	/*
	 * recalculate the amount of metadata blocks to reserve
	 * in order to allocate nrblocks
	 * worse case is one extent per block
	 */
repeat:
	spin_lock(&ei->i_block_reservation_lock);
	/*
	 * ext4_calc_metadata_amount() has side effects, which we have
	 * to be prepared undo if we fail to claim space.
	 */
	save_len = ei->i_da_metadata_calc_len;
	save_last_lblock = ei->i_da_metadata_calc_last_lblock;
	md_needed = EXT4_NUM_B2C(sbi,
				 ext4_calc_metadata_amount(inode, lblock));
	trace_ext4_da_reserve_space(inode, md_needed);

	/*
	 * We do still charge estimated metadata to the sb though;
	 * we cannot afford to run out of free blocks.
	 */
	if (ext4_claim_free_clusters(sbi, md_needed, 0)) {
		ei->i_da_metadata_calc_len = save_len;
		ei->i_da_metadata_calc_last_lblock = save_last_lblock;
		spin_unlock(&ei->i_block_reservation_lock);
		if (ext4_should_retry_alloc(inode->i_sb, &retries)) {
			cond_resched();
			goto repeat;
		}
		return -ENOSPC;
	}
	ei->i_reserved_meta_blocks += md_needed;
	spin_unlock(&ei->i_block_reservation_lock);

	return 0;       /* success */
}

1249
/*
1250
 * Reserve a single cluster located at lblock
1251
 */
1252
static int ext4_da_reserve_space(struct inode *inode, ext4_lblk_t lblock)
1253
{
A
Aneesh Kumar K.V 已提交
1254
	int retries = 0;
1255
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
1256
	struct ext4_inode_info *ei = EXT4_I(inode);
1257
	unsigned int md_needed;
1258
	int ret;
1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269
	ext4_lblk_t save_last_lblock;
	int save_len;

	/*
	 * We will charge metadata quota at writeout time; this saves
	 * us from metadata over-estimation, though we may go over by
	 * a small amount in the end.  Here we just reserve for data.
	 */
	ret = dquot_reserve_block(inode, EXT4_C2B(sbi, 1));
	if (ret)
		return ret;
1270 1271 1272 1273 1274 1275

	/*
	 * recalculate the amount of metadata blocks to reserve
	 * in order to allocate nrblocks
	 * worse case is one extent per block
	 */
A
Aneesh Kumar K.V 已提交
1276
repeat:
1277
	spin_lock(&ei->i_block_reservation_lock);
1278 1279 1280 1281 1282 1283
	/*
	 * ext4_calc_metadata_amount() has side effects, which we have
	 * to be prepared undo if we fail to claim space.
	 */
	save_len = ei->i_da_metadata_calc_len;
	save_last_lblock = ei->i_da_metadata_calc_last_lblock;
1284 1285
	md_needed = EXT4_NUM_B2C(sbi,
				 ext4_calc_metadata_amount(inode, lblock));
1286
	trace_ext4_da_reserve_space(inode, md_needed);
1287

1288 1289 1290 1291
	/*
	 * We do still charge estimated metadata to the sb though;
	 * we cannot afford to run out of free blocks.
	 */
1292
	if (ext4_claim_free_clusters(sbi, md_needed + 1, 0)) {
1293 1294 1295
		ei->i_da_metadata_calc_len = save_len;
		ei->i_da_metadata_calc_last_lblock = save_last_lblock;
		spin_unlock(&ei->i_block_reservation_lock);
A
Aneesh Kumar K.V 已提交
1296
		if (ext4_should_retry_alloc(inode->i_sb, &retries)) {
L
Lukas Czerner 已提交
1297
			cond_resched();
A
Aneesh Kumar K.V 已提交
1298 1299
			goto repeat;
		}
1300
		dquot_release_reservation_block(inode, EXT4_C2B(sbi, 1));
1301 1302
		return -ENOSPC;
	}
1303
	ei->i_reserved_data_blocks++;
1304 1305
	ei->i_reserved_meta_blocks += md_needed;
	spin_unlock(&ei->i_block_reservation_lock);
1306

1307 1308 1309
	return 0;       /* success */
}

1310
static void ext4_da_release_space(struct inode *inode, int to_free)
1311 1312
{
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
1313
	struct ext4_inode_info *ei = EXT4_I(inode);
1314

1315 1316 1317
	if (!to_free)
		return;		/* Nothing to release, exit */

1318
	spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
1319

L
Li Zefan 已提交
1320
	trace_ext4_da_release_space(inode, to_free);
1321
	if (unlikely(to_free > ei->i_reserved_data_blocks)) {
1322
		/*
1323 1324 1325 1326
		 * if there aren't enough reserved blocks, then the
		 * counter is messed up somewhere.  Since this
		 * function is called from invalidate page, it's
		 * harmless to return without any action.
1327
		 */
1328
		ext4_warning(inode->i_sb, "ext4_da_release_space: "
1329
			 "ino %lu, to_free %d with only %d reserved "
1330
			 "data blocks", inode->i_ino, to_free,
1331 1332 1333
			 ei->i_reserved_data_blocks);
		WARN_ON(1);
		to_free = ei->i_reserved_data_blocks;
1334
	}
1335
	ei->i_reserved_data_blocks -= to_free;
1336

1337 1338 1339 1340 1341
	if (ei->i_reserved_data_blocks == 0) {
		/*
		 * We can release all of the reserved metadata blocks
		 * only when we have written all of the delayed
		 * allocation blocks.
1342 1343
		 * Note that in case of bigalloc, i_reserved_meta_blocks,
		 * i_reserved_data_blocks, etc. refer to number of clusters.
1344
		 */
1345
		percpu_counter_sub(&sbi->s_dirtyclusters_counter,
1346
				   ei->i_reserved_meta_blocks);
1347
		ei->i_reserved_meta_blocks = 0;
1348
		ei->i_da_metadata_calc_len = 0;
1349
	}
1350

1351
	/* update fs dirty data blocks counter */
1352
	percpu_counter_sub(&sbi->s_dirtyclusters_counter, to_free);
1353 1354

	spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
1355

1356
	dquot_release_reservation_block(inode, EXT4_C2B(sbi, to_free));
1357 1358 1359
}

static void ext4_da_page_release_reservation(struct page *page,
1360 1361
					     unsigned int offset,
					     unsigned int length)
1362 1363 1364 1365
{
	int to_release = 0;
	struct buffer_head *head, *bh;
	unsigned int curr_off = 0;
1366 1367
	struct inode *inode = page->mapping->host;
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
1368
	unsigned int stop = offset + length;
1369
	int num_clusters;
1370
	ext4_fsblk_t lblk;
1371

1372 1373
	BUG_ON(stop > PAGE_CACHE_SIZE || stop < length);

1374 1375 1376 1377 1378
	head = page_buffers(page);
	bh = head;
	do {
		unsigned int next_off = curr_off + bh->b_size;

1379 1380 1381
		if (next_off > stop)
			break;

1382 1383 1384 1385 1386 1387
		if ((offset <= curr_off) && (buffer_delay(bh))) {
			to_release++;
			clear_buffer_delay(bh);
		}
		curr_off = next_off;
	} while ((bh = bh->b_this_page) != head);
1388

1389 1390 1391 1392 1393
	if (to_release) {
		lblk = page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
		ext4_es_remove_extent(inode, lblk, to_release);
	}

1394 1395 1396 1397 1398 1399 1400
	/* If we have released all the blocks belonging to a cluster, then we
	 * need to release the reserved space for that cluster. */
	num_clusters = EXT4_NUM_B2C(sbi, to_release);
	while (num_clusters > 0) {
		lblk = (page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits)) +
			((num_clusters - 1) << sbi->s_cluster_bits);
		if (sbi->s_cluster_ratio == 1 ||
1401
		    !ext4_find_delalloc_cluster(inode, lblk))
1402 1403 1404 1405
			ext4_da_release_space(inode, 1);

		num_clusters--;
	}
1406
}
1407

1408 1409 1410 1411
/*
 * Delayed allocation stuff
 */

J
Jan Kara 已提交
1412 1413 1414
struct mpage_da_data {
	struct inode *inode;
	struct writeback_control *wbc;
1415

J
Jan Kara 已提交
1416 1417 1418
	pgoff_t first_page;	/* The first page to write */
	pgoff_t next_page;	/* Current page to examine */
	pgoff_t last_page;	/* Last page to examine */
1419
	/*
J
Jan Kara 已提交
1420 1421 1422
	 * Extent to map - this can be after first_page because that can be
	 * fully mapped. We somewhat abuse m_flags to store whether the extent
	 * is delalloc or unwritten.
1423
	 */
J
Jan Kara 已提交
1424 1425 1426
	struct ext4_map_blocks map;
	struct ext4_io_submit io_submit;	/* IO submission data */
};
1427

J
Jan Kara 已提交
1428 1429
static void mpage_release_unused_pages(struct mpage_da_data *mpd,
				       bool invalidate)
1430 1431 1432 1433 1434 1435
{
	int nr_pages, i;
	pgoff_t index, end;
	struct pagevec pvec;
	struct inode *inode = mpd->inode;
	struct address_space *mapping = inode->i_mapping;
J
Jan Kara 已提交
1436 1437 1438 1439

	/* This is necessary when next_page == 0. */
	if (mpd->first_page >= mpd->next_page)
		return;
1440

1441 1442
	index = mpd->first_page;
	end   = mpd->next_page - 1;
J
Jan Kara 已提交
1443 1444 1445 1446 1447 1448
	if (invalidate) {
		ext4_lblk_t start, last;
		start = index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
		last = end << (PAGE_CACHE_SHIFT - inode->i_blkbits);
		ext4_es_remove_extent(inode, start, last - start + 1);
	}
1449

1450
	pagevec_init(&pvec, 0);
1451 1452 1453 1454 1455 1456
	while (index <= end) {
		nr_pages = pagevec_lookup(&pvec, mapping, index, PAGEVEC_SIZE);
		if (nr_pages == 0)
			break;
		for (i = 0; i < nr_pages; i++) {
			struct page *page = pvec.pages[i];
1457
			if (page->index > end)
1458 1459 1460
				break;
			BUG_ON(!PageLocked(page));
			BUG_ON(PageWriteback(page));
J
Jan Kara 已提交
1461 1462 1463 1464
			if (invalidate) {
				block_invalidatepage(page, 0, PAGE_CACHE_SIZE);
				ClearPageUptodate(page);
			}
1465 1466
			unlock_page(page);
		}
1467 1468
		index = pvec.pages[nr_pages - 1]->index + 1;
		pagevec_release(&pvec);
1469 1470 1471
	}
}

1472 1473 1474
static void ext4_print_free_blocks(struct inode *inode)
{
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
1475
	struct super_block *sb = inode->i_sb;
1476
	struct ext4_inode_info *ei = EXT4_I(inode);
1477 1478

	ext4_msg(sb, KERN_CRIT, "Total free blocks count %lld",
1479
	       EXT4_C2B(EXT4_SB(inode->i_sb),
1480
			ext4_count_free_clusters(sb)));
1481 1482
	ext4_msg(sb, KERN_CRIT, "Free/Dirty block details");
	ext4_msg(sb, KERN_CRIT, "free_blocks=%lld",
1483
	       (long long) EXT4_C2B(EXT4_SB(sb),
1484
		percpu_counter_sum(&sbi->s_freeclusters_counter)));
1485
	ext4_msg(sb, KERN_CRIT, "dirty_blocks=%lld",
1486
	       (long long) EXT4_C2B(EXT4_SB(sb),
1487
		percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
1488 1489
	ext4_msg(sb, KERN_CRIT, "Block reservation details");
	ext4_msg(sb, KERN_CRIT, "i_reserved_data_blocks=%u",
1490
		 ei->i_reserved_data_blocks);
1491
	ext4_msg(sb, KERN_CRIT, "i_reserved_meta_blocks=%u",
1492 1493 1494
	       ei->i_reserved_meta_blocks);
	ext4_msg(sb, KERN_CRIT, "i_allocated_meta_blocks=%u",
	       ei->i_allocated_meta_blocks);
1495 1496 1497
	return;
}

1498
static int ext4_bh_delay_or_unwritten(handle_t *handle, struct buffer_head *bh)
1499
{
1500
	return (buffer_delay(bh) || buffer_unwritten(bh)) && buffer_dirty(bh);
1501 1502
}

1503 1504 1505 1506 1507 1508 1509 1510 1511 1512
/*
 * This function is grabs code from the very beginning of
 * ext4_map_blocks, but assumes that the caller is from delayed write
 * time. This function looks up the requested blocks and sets the
 * buffer delay bit under the protection of i_data_sem.
 */
static int ext4_da_map_blocks(struct inode *inode, sector_t iblock,
			      struct ext4_map_blocks *map,
			      struct buffer_head *bh)
{
1513
	struct extent_status es;
1514 1515
	int retval;
	sector_t invalid_block = ~((sector_t) 0xffff);
1516 1517 1518 1519 1520
#ifdef ES_AGGRESSIVE_TEST
	struct ext4_map_blocks orig_map;

	memcpy(&orig_map, map, sizeof(*map));
#endif
1521 1522 1523 1524 1525 1526 1527 1528

	if (invalid_block < ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es))
		invalid_block = ~0;

	map->m_flags = 0;
	ext_debug("ext4_da_map_blocks(): inode %lu, max_blocks %u,"
		  "logical block %lu\n", inode->i_ino, map->m_len,
		  (unsigned long) map->m_lblk);
1529 1530 1531

	/* Lookup extent status tree firstly */
	if (ext4_es_lookup_extent(inode, iblock, &es)) {
1532
		ext4_es_lru_add(inode);
1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561
		if (ext4_es_is_hole(&es)) {
			retval = 0;
			down_read((&EXT4_I(inode)->i_data_sem));
			goto add_delayed;
		}

		/*
		 * Delayed extent could be allocated by fallocate.
		 * So we need to check it.
		 */
		if (ext4_es_is_delayed(&es) && !ext4_es_is_unwritten(&es)) {
			map_bh(bh, inode->i_sb, invalid_block);
			set_buffer_new(bh);
			set_buffer_delay(bh);
			return 0;
		}

		map->m_pblk = ext4_es_pblock(&es) + iblock - es.es_lblk;
		retval = es.es_len - (iblock - es.es_lblk);
		if (retval > map->m_len)
			retval = map->m_len;
		map->m_len = retval;
		if (ext4_es_is_written(&es))
			map->m_flags |= EXT4_MAP_MAPPED;
		else if (ext4_es_is_unwritten(&es))
			map->m_flags |= EXT4_MAP_UNWRITTEN;
		else
			BUG_ON(1);

1562 1563 1564
#ifdef ES_AGGRESSIVE_TEST
		ext4_map_blocks_es_recheck(NULL, inode, map, &orig_map, 0);
#endif
1565 1566 1567
		return retval;
	}

1568 1569 1570 1571 1572
	/*
	 * Try to see if we can get the block without requesting a new
	 * file system block.
	 */
	down_read((&EXT4_I(inode)->i_data_sem));
1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585
	if (ext4_has_inline_data(inode)) {
		/*
		 * We will soon create blocks for this page, and let
		 * us pretend as if the blocks aren't allocated yet.
		 * In case of clusters, we have to handle the work
		 * of mapping from cluster so that the reserved space
		 * is calculated properly.
		 */
		if ((EXT4_SB(inode->i_sb)->s_cluster_ratio > 1) &&
		    ext4_find_delalloc_cluster(inode, map->m_lblk))
			map->m_flags |= EXT4_MAP_FROM_CLUSTER;
		retval = 0;
	} else if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
1586 1587
		retval = ext4_ext_map_blocks(NULL, inode, map,
					     EXT4_GET_BLOCKS_NO_PUT_HOLE);
1588
	else
1589 1590
		retval = ext4_ind_map_blocks(NULL, inode, map,
					     EXT4_GET_BLOCKS_NO_PUT_HOLE);
1591

1592
add_delayed:
1593
	if (retval == 0) {
1594
		int ret;
1595 1596 1597 1598
		/*
		 * XXX: __block_prepare_write() unmaps passed block,
		 * is it OK?
		 */
1599 1600 1601 1602 1603
		/*
		 * If the block was allocated from previously allocated cluster,
		 * then we don't need to reserve it again. However we still need
		 * to reserve metadata for every block we're going to write.
		 */
1604
		if (!(map->m_flags & EXT4_MAP_FROM_CLUSTER)) {
1605 1606
			ret = ext4_da_reserve_space(inode, iblock);
			if (ret) {
1607
				/* not enough space to reserve */
1608
				retval = ret;
1609
				goto out_unlock;
1610
			}
1611 1612 1613 1614 1615 1616 1617
		} else {
			ret = ext4_da_reserve_metadata(inode, iblock);
			if (ret) {
				/* not enough space to reserve */
				retval = ret;
				goto out_unlock;
			}
1618 1619
		}

1620 1621 1622 1623
		ret = ext4_es_insert_extent(inode, map->m_lblk, map->m_len,
					    ~0, EXTENT_STATUS_DELAYED);
		if (ret) {
			retval = ret;
1624
			goto out_unlock;
1625
		}
1626

1627 1628 1629 1630 1631 1632 1633 1634
		/* Clear EXT4_MAP_FROM_CLUSTER flag since its purpose is served
		 * and it should not appear on the bh->b_state.
		 */
		map->m_flags &= ~EXT4_MAP_FROM_CLUSTER;

		map_bh(bh, inode->i_sb, invalid_block);
		set_buffer_new(bh);
		set_buffer_delay(bh);
1635 1636
	} else if (retval > 0) {
		int ret;
1637
		unsigned int status;
1638

1639 1640 1641 1642 1643 1644
		if (unlikely(retval != map->m_len)) {
			ext4_warning(inode->i_sb,
				     "ES len assertion failed for inode "
				     "%lu: retval %d != map->m_len %d",
				     inode->i_ino, retval, map->m_len);
			WARN_ON(1);
1645 1646
		}

1647 1648 1649 1650 1651 1652
		status = map->m_flags & EXT4_MAP_UNWRITTEN ?
				EXTENT_STATUS_UNWRITTEN : EXTENT_STATUS_WRITTEN;
		ret = ext4_es_insert_extent(inode, map->m_lblk, map->m_len,
					    map->m_pblk, status);
		if (ret != 0)
			retval = ret;
1653 1654 1655 1656 1657 1658 1659 1660
	}

out_unlock:
	up_read((&EXT4_I(inode)->i_data_sem));

	return retval;
}

1661
/*
1662 1663 1664
 * This is a special get_blocks_t callback which is used by
 * ext4_da_write_begin().  It will either return mapped block or
 * reserve space for a single block.
1665 1666 1667 1668 1669 1670 1671
 *
 * For delayed buffer_head we have BH_Mapped, BH_New, BH_Delay set.
 * We also have b_blocknr = -1 and b_bdev initialized properly
 *
 * For unwritten buffer_head we have BH_Mapped, BH_New, BH_Unwritten set.
 * We also have b_blocknr = physicalblock mapping unwritten extent and b_bdev
 * initialized properly.
1672
 */
1673 1674
int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
			   struct buffer_head *bh, int create)
1675
{
1676
	struct ext4_map_blocks map;
1677 1678 1679
	int ret = 0;

	BUG_ON(create == 0);
1680 1681 1682 1683
	BUG_ON(bh->b_size != inode->i_sb->s_blocksize);

	map.m_lblk = iblock;
	map.m_len = 1;
1684 1685 1686 1687 1688 1689

	/*
	 * first, we need to know whether the block is allocated already
	 * preallocated blocks are unmapped but should treated
	 * the same as allocated blocks.
	 */
1690 1691
	ret = ext4_da_map_blocks(inode, iblock, &map, bh);
	if (ret <= 0)
1692
		return ret;
1693

1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704
	map_bh(bh, inode->i_sb, map.m_pblk);
	bh->b_state = (bh->b_state & ~EXT4_MAP_FLAGS) | map.m_flags;

	if (buffer_unwritten(bh)) {
		/* A delayed write to unwritten bh should be marked
		 * new and mapped.  Mapped ensures that we don't do
		 * get_block multiple times when we write to the same
		 * offset and new ensures that we do proper zero out
		 * for partial write.
		 */
		set_buffer_new(bh);
1705
		set_buffer_mapped(bh);
1706 1707
	}
	return 0;
1708
}
1709

1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726
static int bget_one(handle_t *handle, struct buffer_head *bh)
{
	get_bh(bh);
	return 0;
}

static int bput_one(handle_t *handle, struct buffer_head *bh)
{
	put_bh(bh);
	return 0;
}

static int __ext4_journalled_writepage(struct page *page,
				       unsigned int len)
{
	struct address_space *mapping = page->mapping;
	struct inode *inode = mapping->host;
1727
	struct buffer_head *page_bufs = NULL;
1728
	handle_t *handle = NULL;
1729 1730 1731
	int ret = 0, err = 0;
	int inline_data = ext4_has_inline_data(inode);
	struct buffer_head *inode_bh = NULL;
1732

1733
	ClearPageChecked(page);
1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749

	if (inline_data) {
		BUG_ON(page->index != 0);
		BUG_ON(len > ext4_get_max_inline_size(inode));
		inode_bh = ext4_journalled_write_inline_data(inode, len, page);
		if (inode_bh == NULL)
			goto out;
	} else {
		page_bufs = page_buffers(page);
		if (!page_bufs) {
			BUG();
			goto out;
		}
		ext4_walk_page_buffers(handle, page_bufs, 0, len,
				       NULL, bget_one);
	}
1750 1751 1752 1753
	/* As soon as we unlock the page, it can go away, but we have
	 * references to buffers so we are safe */
	unlock_page(page);

1754 1755
	handle = ext4_journal_start(inode, EXT4_HT_WRITE_PAGE,
				    ext4_writepage_trans_blocks(inode));
1756 1757 1758 1759 1760
	if (IS_ERR(handle)) {
		ret = PTR_ERR(handle);
		goto out;
	}

1761 1762
	BUG_ON(!ext4_handle_valid(handle));

1763 1764
	if (inline_data) {
		ret = ext4_journal_get_write_access(handle, inode_bh);
1765

1766 1767 1768 1769 1770 1771 1772 1773 1774
		err = ext4_handle_dirty_metadata(handle, inode, inode_bh);

	} else {
		ret = ext4_walk_page_buffers(handle, page_bufs, 0, len, NULL,
					     do_journal_get_write_access);

		err = ext4_walk_page_buffers(handle, page_bufs, 0, len, NULL,
					     write_end_fn);
	}
1775 1776
	if (ret == 0)
		ret = err;
1777
	EXT4_I(inode)->i_datasync_tid = handle->h_transaction->t_tid;
1778 1779 1780 1781
	err = ext4_journal_stop(handle);
	if (!ret)
		ret = err;

1782 1783 1784
	if (!ext4_has_inline_data(inode))
		ext4_walk_page_buffers(handle, page_bufs, 0, len,
				       NULL, bput_one);
1785
	ext4_set_inode_state(inode, EXT4_STATE_JDATA);
1786
out:
1787
	brelse(inode_bh);
1788 1789 1790
	return ret;
}

1791
/*
1792 1793 1794 1795
 * Note that we don't need to start a transaction unless we're journaling data
 * because we should have holes filled from ext4_page_mkwrite(). We even don't
 * need to file the inode to the transaction's list in ordered mode because if
 * we are writing back data added by write(), the inode is already there and if
L
Lucas De Marchi 已提交
1796
 * we are writing back data modified via mmap(), no one guarantees in which
1797 1798 1799 1800
 * transaction the data will hit the disk. In case we are journaling data, we
 * cannot start transaction directly because transaction start ranks above page
 * lock so we have to do some magic.
 *
1801
 * This function can get called via...
1802
 *   - ext4_writepages after taking page lock (have journal handle)
1803
 *   - journal_submit_inode_data_buffers (no journal handle)
1804
 *   - shrink_page_list via the kswapd/direct reclaim (no journal handle)
1805
 *   - grab_page_cache when doing write_begin (have journal handle)
1806 1807 1808 1809 1810 1811 1812 1813 1814
 *
 * We don't do any block allocation in this function. If we have page with
 * multiple blocks we need to write those buffer_heads that are mapped. This
 * is important for mmaped based write. So if we do with blocksize 1K
 * truncate(f, 1024);
 * a = mmap(f, 0, 4096);
 * a[0] = 'a';
 * truncate(f, 4096);
 * we have in the page first buffer_head mapped via page_mkwrite call back
1815
 * but other buffer_heads would be unmapped but dirty (dirty done via the
1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830
 * do_wp_page). So writepage should write the first block. If we modify
 * the mmap area beyond 1024 we will again get a page_fault and the
 * page_mkwrite callback will do the block allocation and mark the
 * buffer_heads mapped.
 *
 * We redirty the page if we have any buffer_heads that is either delay or
 * unwritten in the page.
 *
 * We can get recursively called as show below.
 *
 *	ext4_writepage() -> kmalloc() -> __alloc_pages() -> page_launder() ->
 *		ext4_writepage()
 *
 * But since we don't do any block allocation we should not deadlock.
 * Page also have the dirty flag cleared so we don't get recurive page_lock.
1831
 */
1832
static int ext4_writepage(struct page *page,
1833
			  struct writeback_control *wbc)
1834
{
1835
	int ret = 0;
1836
	loff_t size;
1837
	unsigned int len;
1838
	struct buffer_head *page_bufs = NULL;
1839
	struct inode *inode = page->mapping->host;
1840
	struct ext4_io_submit io_submit;
1841

L
Lukas Czerner 已提交
1842
	trace_ext4_writepage(page);
1843 1844 1845 1846 1847
	size = i_size_read(inode);
	if (page->index == size >> PAGE_CACHE_SHIFT)
		len = size & ~PAGE_CACHE_MASK;
	else
		len = PAGE_CACHE_SIZE;
1848

T
Theodore Ts'o 已提交
1849 1850
	page_bufs = page_buffers(page);
	/*
1851 1852 1853 1854 1855
	 * We cannot do block allocation or other extent handling in this
	 * function. If there are buffers needing that, we have to redirty
	 * the page. But we may reach here when we do a journal commit via
	 * journal_submit_inode_data_buffers() and in that case we must write
	 * allocated buffers to achieve data=ordered mode guarantees.
T
Theodore Ts'o 已提交
1856
	 */
1857 1858
	if (ext4_walk_page_buffers(NULL, page_bufs, 0, len, NULL,
				   ext4_bh_delay_or_unwritten)) {
1859
		redirty_page_for_writepage(wbc, page);
1860 1861 1862 1863 1864 1865 1866 1867
		if (current->flags & PF_MEMALLOC) {
			/*
			 * For memory cleaning there's no point in writing only
			 * some buffers. So just bail out. Warn if we came here
			 * from direct reclaim.
			 */
			WARN_ON_ONCE((current->flags & (PF_MEMALLOC|PF_KSWAPD))
							== PF_MEMALLOC);
1868 1869 1870
			unlock_page(page);
			return 0;
		}
T
Theodore Ts'o 已提交
1871
	}
1872

1873
	if (PageChecked(page) && ext4_should_journal_data(inode))
1874 1875 1876 1877
		/*
		 * It's mmapped pagecache.  Add buffers and journal it.  There
		 * doesn't seem much point in redirtying the page here.
		 */
1878
		return __ext4_journalled_writepage(page, len);
1879

J
Jan Kara 已提交
1880 1881 1882 1883 1884 1885 1886
	ext4_io_submit_init(&io_submit, wbc);
	io_submit.io_end = ext4_init_io_end(inode, GFP_NOFS);
	if (!io_submit.io_end) {
		redirty_page_for_writepage(wbc, page);
		unlock_page(page);
		return -ENOMEM;
	}
1887 1888
	ret = ext4_bio_write_page(&io_submit, page, len, wbc);
	ext4_io_submit(&io_submit);
J
Jan Kara 已提交
1889 1890
	/* Drop io_end reference we got from init */
	ext4_put_io_end_defer(io_submit.io_end);
1891 1892 1893
	return ret;
}

1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913
static int mpage_submit_page(struct mpage_da_data *mpd, struct page *page)
{
	int len;
	loff_t size = i_size_read(mpd->inode);
	int err;

	BUG_ON(page->index != mpd->first_page);
	if (page->index == size >> PAGE_CACHE_SHIFT)
		len = size & ~PAGE_CACHE_MASK;
	else
		len = PAGE_CACHE_SIZE;
	clear_page_dirty_for_io(page);
	err = ext4_bio_write_page(&mpd->io_submit, page, len, mpd->wbc);
	if (!err)
		mpd->wbc->nr_to_write--;
	mpd->first_page++;

	return err;
}

J
Jan Kara 已提交
1914 1915
#define BH_FLAGS ((1 << BH_Unwritten) | (1 << BH_Delay))

1916
/*
1917 1918 1919
 * mballoc gives us at most this number of blocks...
 * XXX: That seems to be only a limitation of ext4_mb_normalize_request().
 * The rest of mballoc seems to handle chunks upto full group size.
1920
 */
1921
#define MAX_WRITEPAGES_EXTENT_LEN 2048
1922

J
Jan Kara 已提交
1923 1924 1925 1926 1927
/*
 * mpage_add_bh_to_extent - try to add bh to extent of blocks to map
 *
 * @mpd - extent of blocks
 * @lblk - logical number of the block in the file
1928
 * @bh - buffer head we want to add to the extent
J
Jan Kara 已提交
1929
 *
1930 1931 1932 1933 1934 1935
 * The function is used to collect contig. blocks in the same state. If the
 * buffer doesn't require mapping for writeback and we haven't started the
 * extent of buffers to map yet, the function returns 'true' immediately - the
 * caller can write the buffer right away. Otherwise the function returns true
 * if the block has been added to the extent, false if the block couldn't be
 * added.
J
Jan Kara 已提交
1936
 */
1937 1938
static bool mpage_add_bh_to_extent(struct mpage_da_data *mpd, ext4_lblk_t lblk,
				   struct buffer_head *bh)
J
Jan Kara 已提交
1939 1940 1941
{
	struct ext4_map_blocks *map = &mpd->map;

1942 1943 1944 1945 1946 1947 1948 1949
	/* Buffer that doesn't need mapping for writeback? */
	if (!buffer_dirty(bh) || !buffer_mapped(bh) ||
	    (!buffer_delay(bh) && !buffer_unwritten(bh))) {
		/* So far no extent to map => we write the buffer right away */
		if (map->m_len == 0)
			return true;
		return false;
	}
J
Jan Kara 已提交
1950 1951 1952 1953 1954

	/* First block in the extent? */
	if (map->m_len == 0) {
		map->m_lblk = lblk;
		map->m_len = 1;
1955 1956
		map->m_flags = bh->b_state & BH_FLAGS;
		return true;
J
Jan Kara 已提交
1957 1958
	}

1959 1960 1961 1962
	/* Don't go larger than mballoc is willing to allocate */
	if (map->m_len >= MAX_WRITEPAGES_EXTENT_LEN)
		return false;

J
Jan Kara 已提交
1963 1964
	/* Can we merge the block to our big extent? */
	if (lblk == map->m_lblk + map->m_len &&
1965
	    (bh->b_state & BH_FLAGS) == map->m_flags) {
J
Jan Kara 已提交
1966
		map->m_len++;
1967
		return true;
J
Jan Kara 已提交
1968
	}
1969
	return false;
J
Jan Kara 已提交
1970 1971
}

1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991
/*
 * mpage_process_page_bufs - submit page buffers for IO or add them to extent
 *
 * @mpd - extent of blocks for mapping
 * @head - the first buffer in the page
 * @bh - buffer we should start processing from
 * @lblk - logical number of the block in the file corresponding to @bh
 *
 * Walk through page buffers from @bh upto @head (exclusive) and either submit
 * the page for IO if all buffers in this page were mapped and there's no
 * accumulated extent of buffers to map or add buffers in the page to the
 * extent of buffers to map. The function returns 1 if the caller can continue
 * by processing the next page, 0 if it should stop adding buffers to the
 * extent to map because we cannot extend it anymore. It can also return value
 * < 0 in case of error during IO submission.
 */
static int mpage_process_page_bufs(struct mpage_da_data *mpd,
				   struct buffer_head *head,
				   struct buffer_head *bh,
				   ext4_lblk_t lblk)
J
Jan Kara 已提交
1992 1993
{
	struct inode *inode = mpd->inode;
1994
	int err;
J
Jan Kara 已提交
1995 1996 1997 1998 1999 2000
	ext4_lblk_t blocks = (i_size_read(inode) + (1 << inode->i_blkbits) - 1)
							>> inode->i_blkbits;

	do {
		BUG_ON(buffer_locked(bh));

2001
		if (lblk >= blocks || !mpage_add_bh_to_extent(mpd, lblk, bh)) {
J
Jan Kara 已提交
2002 2003
			/* Found extent to map? */
			if (mpd->map.m_len)
2004
				return 0;
2005
			/* Everything mapped so far and we hit EOF */
2006
			break;
J
Jan Kara 已提交
2007 2008
		}
	} while (lblk++, (bh = bh->b_this_page) != head);
2009 2010 2011 2012 2013 2014 2015
	/* So far everything mapped? Submit the page for IO. */
	if (mpd->map.m_len == 0) {
		err = mpage_submit_page(mpd, head->b_page);
		if (err < 0)
			return err;
	}
	return lblk < blocks;
J
Jan Kara 已提交
2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072
}

/*
 * mpage_map_buffers - update buffers corresponding to changed extent and
 *		       submit fully mapped pages for IO
 *
 * @mpd - description of extent to map, on return next extent to map
 *
 * Scan buffers corresponding to changed extent (we expect corresponding pages
 * to be already locked) and update buffer state according to new extent state.
 * We map delalloc buffers to their physical location, clear unwritten bits,
 * and mark buffers as uninit when we perform writes to uninitialized extents
 * and do extent conversion after IO is finished. If the last page is not fully
 * mapped, we update @map to the next extent in the last page that needs
 * mapping. Otherwise we submit the page for IO.
 */
static int mpage_map_and_submit_buffers(struct mpage_da_data *mpd)
{
	struct pagevec pvec;
	int nr_pages, i;
	struct inode *inode = mpd->inode;
	struct buffer_head *head, *bh;
	int bpp_bits = PAGE_CACHE_SHIFT - inode->i_blkbits;
	pgoff_t start, end;
	ext4_lblk_t lblk;
	sector_t pblock;
	int err;

	start = mpd->map.m_lblk >> bpp_bits;
	end = (mpd->map.m_lblk + mpd->map.m_len - 1) >> bpp_bits;
	lblk = start << bpp_bits;
	pblock = mpd->map.m_pblk;

	pagevec_init(&pvec, 0);
	while (start <= end) {
		nr_pages = pagevec_lookup(&pvec, inode->i_mapping, start,
					  PAGEVEC_SIZE);
		if (nr_pages == 0)
			break;
		for (i = 0; i < nr_pages; i++) {
			struct page *page = pvec.pages[i];

			if (page->index > end)
				break;
			/* Upto 'end' pages must be contiguous */
			BUG_ON(page->index != start);
			bh = head = page_buffers(page);
			do {
				if (lblk < mpd->map.m_lblk)
					continue;
				if (lblk >= mpd->map.m_lblk + mpd->map.m_len) {
					/*
					 * Buffer after end of mapped extent.
					 * Find next buffer in the page to map.
					 */
					mpd->map.m_len = 0;
					mpd->map.m_flags = 0;
2073 2074 2075 2076 2077 2078 2079 2080 2081
					/*
					 * FIXME: If dioread_nolock supports
					 * blocksize < pagesize, we need to make
					 * sure we add size mapped so far to
					 * io_end->size as the following call
					 * can submit the page for IO.
					 */
					err = mpage_process_page_bufs(mpd, head,
								      bh, lblk);
J
Jan Kara 已提交
2082
					pagevec_release(&pvec);
2083 2084 2085
					if (err > 0)
						err = 0;
					return err;
J
Jan Kara 已提交
2086 2087 2088 2089 2090 2091
				}
				if (buffer_delay(bh)) {
					clear_buffer_delay(bh);
					bh->b_blocknr = pblock++;
				}
				clear_buffer_unwritten(bh);
2092
			} while (lblk++, (bh = bh->b_this_page) != head);
J
Jan Kara 已提交
2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150

			/*
			 * FIXME: This is going to break if dioread_nolock
			 * supports blocksize < pagesize as we will try to
			 * convert potentially unmapped parts of inode.
			 */
			mpd->io_submit.io_end->size += PAGE_CACHE_SIZE;
			/* Page fully mapped - let IO run! */
			err = mpage_submit_page(mpd, page);
			if (err < 0) {
				pagevec_release(&pvec);
				return err;
			}
			start++;
		}
		pagevec_release(&pvec);
	}
	/* Extent fully mapped and matches with page boundary. We are done. */
	mpd->map.m_len = 0;
	mpd->map.m_flags = 0;
	return 0;
}

static int mpage_map_one_extent(handle_t *handle, struct mpage_da_data *mpd)
{
	struct inode *inode = mpd->inode;
	struct ext4_map_blocks *map = &mpd->map;
	int get_blocks_flags;
	int err;

	trace_ext4_da_write_pages_extent(inode, map);
	/*
	 * Call ext4_map_blocks() to allocate any delayed allocation blocks, or
	 * to convert an uninitialized extent to be initialized (in the case
	 * where we have written into one or more preallocated blocks).  It is
	 * possible that we're going to need more metadata blocks than
	 * previously reserved. However we must not fail because we're in
	 * writeback and there is nothing we can do about it so it might result
	 * in data loss.  So use reserved blocks to allocate metadata if
	 * possible.
	 *
	 * We pass in the magic EXT4_GET_BLOCKS_DELALLOC_RESERVE if the blocks
	 * in question are delalloc blocks.  This affects functions in many
	 * different parts of the allocation call path.  This flag exists
	 * primarily because we don't want to change *many* call functions, so
	 * ext4_map_blocks() will set the EXT4_STATE_DELALLOC_RESERVED flag
	 * once the inode's allocation semaphore is taken.
	 */
	get_blocks_flags = EXT4_GET_BLOCKS_CREATE |
			   EXT4_GET_BLOCKS_METADATA_NOFAIL;
	if (ext4_should_dioread_nolock(inode))
		get_blocks_flags |= EXT4_GET_BLOCKS_IO_CREATE_EXT;
	if (map->m_flags & (1 << BH_Delay))
		get_blocks_flags |= EXT4_GET_BLOCKS_DELALLOC_RESERVE;

	err = ext4_map_blocks(handle, inode, map, get_blocks_flags);
	if (err < 0)
		return err;
2151 2152 2153 2154 2155 2156
	if (map->m_flags & EXT4_MAP_UNINIT) {
		if (!mpd->io_submit.io_end->handle &&
		    ext4_handle_valid(handle)) {
			mpd->io_submit.io_end->handle = handle->h_rsv_handle;
			handle->h_rsv_handle = NULL;
		}
J
Jan Kara 已提交
2157
		ext4_set_io_unwritten_flag(inode, mpd->io_submit.io_end);
2158
	}
J
Jan Kara 已提交
2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188

	BUG_ON(map->m_len == 0);
	if (map->m_flags & EXT4_MAP_NEW) {
		struct block_device *bdev = inode->i_sb->s_bdev;
		int i;

		for (i = 0; i < map->m_len; i++)
			unmap_underlying_metadata(bdev, map->m_pblk + i);
	}
	return 0;
}

/*
 * mpage_map_and_submit_extent - map extent starting at mpd->lblk of length
 *				 mpd->len and submit pages underlying it for IO
 *
 * @handle - handle for journal operations
 * @mpd - extent to map
 *
 * The function maps extent starting at mpd->lblk of length mpd->len. If it is
 * delayed, blocks are allocated, if it is unwritten, we may need to convert
 * them to initialized or split the described range from larger unwritten
 * extent. Note that we need not map all the described range since allocation
 * can return less blocks or the range is covered by more unwritten extents. We
 * cannot map more because we are limited by reserved transaction credits. On
 * the other hand we always make sure that the last touched page is fully
 * mapped so that it can be written out (and thus forward progress is
 * guaranteed). After mapping we submit all mapped pages for IO.
 */
static int mpage_map_and_submit_extent(handle_t *handle,
2189 2190
				       struct mpage_da_data *mpd,
				       bool *give_up_on_write)
J
Jan Kara 已提交
2191 2192 2193 2194 2195 2196 2197 2198
{
	struct inode *inode = mpd->inode;
	struct ext4_map_blocks *map = &mpd->map;
	int err;
	loff_t disksize;

	mpd->io_submit.io_end->offset =
				((loff_t)map->m_lblk) << inode->i_blkbits;
2199
	do {
J
Jan Kara 已提交
2200 2201 2202 2203
		err = mpage_map_one_extent(handle, mpd);
		if (err < 0) {
			struct super_block *sb = inode->i_sb;

2204 2205
			if (EXT4_SB(sb)->s_mount_flags & EXT4_MF_FS_ABORTED)
				goto invalidate_dirty_pages;
J
Jan Kara 已提交
2206
			/*
2207 2208 2209
			 * Let the uper layers retry transient errors.
			 * In the case of ENOSPC, if ext4_count_free_blocks()
			 * is non-zero, a commit should free up blocks.
J
Jan Kara 已提交
2210
			 */
2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227
			if ((err == -ENOMEM) ||
			    (err == -ENOSPC && ext4_count_free_clusters(sb)))
				return err;
			ext4_msg(sb, KERN_CRIT,
				 "Delayed block allocation failed for "
				 "inode %lu at logical offset %llu with"
				 " max blocks %u with error %d",
				 inode->i_ino,
				 (unsigned long long)map->m_lblk,
				 (unsigned)map->m_len, -err);
			ext4_msg(sb, KERN_CRIT,
				 "This should not happen!! Data will "
				 "be lost\n");
			if (err == -ENOSPC)
				ext4_print_free_blocks(inode);
		invalidate_dirty_pages:
			*give_up_on_write = true;
J
Jan Kara 已提交
2228 2229 2230 2231 2232 2233 2234 2235 2236
			return err;
		}
		/*
		 * Update buffer state, submit mapped pages, and get us new
		 * extent to map
		 */
		err = mpage_map_and_submit_buffers(mpd);
		if (err < 0)
			return err;
2237
	} while (map->m_len);
J
Jan Kara 已提交
2238 2239 2240 2241 2242 2243

	/* Update on-disk size after IO is submitted */
	disksize = ((loff_t)mpd->first_page) << PAGE_CACHE_SHIFT;
	if (disksize > EXT4_I(inode)->i_disksize) {
		int err2;

2244
		ext4_wb_update_i_disksize(inode, disksize);
J
Jan Kara 已提交
2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255
		err2 = ext4_mark_inode_dirty(handle, inode);
		if (err2)
			ext4_error(inode->i_sb,
				   "Failed to mark inode %lu dirty",
				   inode->i_ino);
		if (!err)
			err = err2;
	}
	return err;
}

2256 2257
/*
 * Calculate the total number of credits to reserve for one writepages
2258
 * iteration. This is called from ext4_writepages(). We map an extent of
2259 2260 2261 2262
 * upto MAX_WRITEPAGES_EXTENT_LEN blocks and then we go on and finish mapping
 * the last partial page. So in total we can map MAX_WRITEPAGES_EXTENT_LEN +
 * bpp - 1 blocks in bpp different extents.
 */
2263 2264
static int ext4_da_writepages_trans_blocks(struct inode *inode)
{
2265
	int bpp = ext4_journal_blocks_per_page(inode);
2266

2267 2268
	return ext4_meta_trans_blocks(inode,
				MAX_WRITEPAGES_EXTENT_LEN + bpp - 1, bpp);
2269
}
2270

2271
/*
J
Jan Kara 已提交
2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287
 * mpage_prepare_extent_to_map - find & lock contiguous range of dirty pages
 * 				 and underlying extent to map
 *
 * @mpd - where to look for pages
 *
 * Walk dirty pages in the mapping. If they are fully mapped, submit them for
 * IO immediately. When we find a page which isn't mapped we start accumulating
 * extent of buffers underlying these pages that needs mapping (formed by
 * either delayed or unwritten buffers). We also lock the pages containing
 * these buffers. The extent found is returned in @mpd structure (starting at
 * mpd->lblk with length mpd->len blocks).
 *
 * Note that this function can attach bios to one io_end structure which are
 * neither logically nor physically contiguous. Although it may seem as an
 * unnecessary complication, it is actually inevitable in blocksize < pagesize
 * case as we need to track IO to all buffers underlying a page in one io_end.
2288
 */
J
Jan Kara 已提交
2289
static int mpage_prepare_extent_to_map(struct mpage_da_data *mpd)
2290
{
J
Jan Kara 已提交
2291 2292 2293 2294 2295 2296 2297 2298 2299 2300
	struct address_space *mapping = mpd->inode->i_mapping;
	struct pagevec pvec;
	unsigned int nr_pages;
	pgoff_t index = mpd->first_page;
	pgoff_t end = mpd->last_page;
	int tag;
	int i, err = 0;
	int blkbits = mpd->inode->i_blkbits;
	ext4_lblk_t lblk;
	struct buffer_head *head;
2301

J
Jan Kara 已提交
2302
	if (mpd->wbc->sync_mode == WB_SYNC_ALL || mpd->wbc->tagged_writepages)
2303 2304 2305 2306
		tag = PAGECACHE_TAG_TOWRITE;
	else
		tag = PAGECACHE_TAG_DIRTY;

J
Jan Kara 已提交
2307 2308 2309
	pagevec_init(&pvec, 0);
	mpd->map.m_len = 0;
	mpd->next_page = index;
2310
	while (index <= end) {
2311
		nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag,
2312 2313
			      min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1);
		if (nr_pages == 0)
J
Jan Kara 已提交
2314
			goto out;
2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325

		for (i = 0; i < nr_pages; i++) {
			struct page *page = pvec.pages[i];

			/*
			 * At this point, the page may be truncated or
			 * invalidated (changing page->mapping to NULL), or
			 * even swizzled back from swapper_space to tmpfs file
			 * mapping. However, page->index will not change
			 * because we have a reference on the page.
			 */
2326 2327
			if (page->index > end)
				goto out;
2328

J
Jan Kara 已提交
2329 2330 2331
			/* If we can't merge this page, we are done. */
			if (mpd->map.m_len > 0 && mpd->next_page != page->index)
				goto out;
2332

2333 2334
			lock_page(page);
			/*
J
Jan Kara 已提交
2335 2336 2337 2338 2339
			 * If the page is no longer dirty, or its mapping no
			 * longer corresponds to inode we are writing (which
			 * means it has been truncated or invalidated), or the
			 * page is already under writeback and we are not doing
			 * a data integrity writeback, skip the page
2340
			 */
2341 2342
			if (!PageDirty(page) ||
			    (PageWriteback(page) &&
J
Jan Kara 已提交
2343
			     (mpd->wbc->sync_mode == WB_SYNC_NONE)) ||
2344
			    unlikely(page->mapping != mapping)) {
2345 2346 2347 2348
				unlock_page(page);
				continue;
			}

2349
			wait_on_page_writeback(page);
2350 2351
			BUG_ON(PageWriteback(page));

J
Jan Kara 已提交
2352
			if (mpd->map.m_len == 0)
2353 2354
				mpd->first_page = page->index;
			mpd->next_page = page->index + 1;
2355
			/* Add all dirty buffers to mpd */
J
Jan Kara 已提交
2356 2357
			lblk = ((ext4_lblk_t)page->index) <<
				(PAGE_CACHE_SHIFT - blkbits);
2358
			head = page_buffers(page);
2359 2360
			err = mpage_process_page_bufs(mpd, head, head, lblk);
			if (err <= 0)
J
Jan Kara 已提交
2361
				goto out;
2362
			err = 0;
J
Jan Kara 已提交
2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375

			/*
			 * Accumulated enough dirty pages? This doesn't apply
			 * to WB_SYNC_ALL mode. For integrity sync we have to
			 * keep going because someone may be concurrently
			 * dirtying pages, and we might have synced a lot of
			 * newly appeared dirty pages, but have not synced all
			 * of the old dirty pages.
			 */
			if (mpd->wbc->sync_mode == WB_SYNC_NONE &&
			    mpd->next_page - mpd->first_page >=
							mpd->wbc->nr_to_write)
				goto out;
2376 2377 2378 2379
		}
		pagevec_release(&pvec);
		cond_resched();
	}
2380
	return 0;
2381 2382
out:
	pagevec_release(&pvec);
J
Jan Kara 已提交
2383
	return err;
2384 2385
}

2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396
static int __writepage(struct page *page, struct writeback_control *wbc,
		       void *data)
{
	struct address_space *mapping = data;
	int ret = ext4_writepage(page, wbc);
	mapping_set_error(mapping, ret);
	return ret;
}

static int ext4_writepages(struct address_space *mapping,
			   struct writeback_control *wbc)
2397
{
J
Jan Kara 已提交
2398 2399
	pgoff_t	writeback_index = 0;
	long nr_to_write = wbc->nr_to_write;
2400
	int range_whole = 0;
J
Jan Kara 已提交
2401
	int cycled = 1;
2402
	handle_t *handle = NULL;
2403
	struct mpage_da_data mpd;
2404
	struct inode *inode = mapping->host;
2405
	int needed_blocks, rsv_blocks = 0, ret = 0;
2406
	struct ext4_sb_info *sbi = EXT4_SB(mapping->host->i_sb);
J
Jan Kara 已提交
2407
	bool done;
S
Shaohua Li 已提交
2408
	struct blk_plug plug;
2409
	bool give_up_on_write = false;
2410

2411
	trace_ext4_writepages(inode, wbc);
2412

2413 2414 2415 2416 2417
	/*
	 * No pages to write? This is mainly a kludge to avoid starting
	 * a transaction for special inodes like journal inode on last iput()
	 * because that could violate lock ordering on umount
	 */
2418
	if (!mapping->nrpages || !mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
2419
		return 0;
2420

2421 2422 2423 2424 2425 2426 2427 2428 2429 2430
	if (ext4_should_journal_data(inode)) {
		struct blk_plug plug;
		int ret;

		blk_start_plug(&plug);
		ret = write_cache_pages(mapping, wbc, __writepage, mapping);
		blk_finish_plug(&plug);
		return ret;
	}

2431 2432 2433 2434
	/*
	 * If the filesystem has aborted, it is read-only, so return
	 * right away instead of dumping stack traces later on that
	 * will obscure the real source of the problem.  We test
2435
	 * EXT4_MF_FS_ABORTED instead of sb->s_flag's MS_RDONLY because
2436
	 * the latter could be true if the filesystem is mounted
2437
	 * read-only, and in that case, ext4_writepages should
2438 2439 2440
	 * *never* be called, so if that ever happens, we would want
	 * the stack trace.
	 */
2441
	if (unlikely(sbi->s_mount_flags & EXT4_MF_FS_ABORTED))
2442 2443
		return -EROFS;

2444 2445 2446 2447 2448 2449 2450 2451
	if (ext4_should_dioread_nolock(inode)) {
		/*
		 * We may need to convert upto one extent per block in
		 * the page and we may dirty the inode.
		 */
		rsv_blocks = 1 + (PAGE_CACHE_SIZE >> inode->i_blkbits);
	}

J
Jan Kara 已提交
2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469
	/*
	 * If we have inline data and arrive here, it means that
	 * we will soon create the block for the 1st page, so
	 * we'd better clear the inline data here.
	 */
	if (ext4_has_inline_data(inode)) {
		/* Just inode will be modified... */
		handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
		if (IS_ERR(handle)) {
			ret = PTR_ERR(handle);
			goto out_writepages;
		}
		BUG_ON(ext4_test_inode_state(inode,
				EXT4_STATE_MAY_INLINE_DATA));
		ext4_destroy_inline_data(handle, inode);
		ext4_journal_stop(handle);
	}

2470 2471
	if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
		range_whole = 1;
2472

2473
	if (wbc->range_cyclic) {
J
Jan Kara 已提交
2474 2475
		writeback_index = mapping->writeback_index;
		if (writeback_index)
2476
			cycled = 0;
J
Jan Kara 已提交
2477 2478
		mpd.first_page = writeback_index;
		mpd.last_page = -1;
2479
	} else {
J
Jan Kara 已提交
2480 2481
		mpd.first_page = wbc->range_start >> PAGE_CACHE_SHIFT;
		mpd.last_page = wbc->range_end >> PAGE_CACHE_SHIFT;
2482
	}
2483

J
Jan Kara 已提交
2484 2485 2486
	mpd.inode = inode;
	mpd.wbc = wbc;
	ext4_io_submit_init(&mpd.io_submit, wbc);
2487
retry:
2488
	if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
J
Jan Kara 已提交
2489 2490
		tag_pages_for_writeback(mapping, mpd.first_page, mpd.last_page);
	done = false;
S
Shaohua Li 已提交
2491
	blk_start_plug(&plug);
J
Jan Kara 已提交
2492 2493 2494 2495 2496 2497 2498
	while (!done && mpd.first_page <= mpd.last_page) {
		/* For each extent of pages we use new io_end */
		mpd.io_submit.io_end = ext4_init_io_end(inode, GFP_KERNEL);
		if (!mpd.io_submit.io_end) {
			ret = -ENOMEM;
			break;
		}
2499 2500

		/*
J
Jan Kara 已提交
2501 2502 2503 2504 2505
		 * We have two constraints: We find one extent to map and we
		 * must always write out whole page (makes a difference when
		 * blocksize < pagesize) so that we don't block on IO when we
		 * try to write out the rest of the page. Journalled mode is
		 * not supported by delalloc.
2506 2507
		 */
		BUG_ON(ext4_should_journal_data(inode));
2508
		needed_blocks = ext4_da_writepages_trans_blocks(inode);
2509

J
Jan Kara 已提交
2510
		/* start a new transaction */
2511 2512
		handle = ext4_journal_start_with_reserve(inode,
				EXT4_HT_WRITE_PAGE, needed_blocks, rsv_blocks);
2513 2514
		if (IS_ERR(handle)) {
			ret = PTR_ERR(handle);
2515
			ext4_msg(inode->i_sb, KERN_CRIT, "%s: jbd2_start: "
2516
			       "%ld pages, ino %lu; err %d", __func__,
2517
				wbc->nr_to_write, inode->i_ino, ret);
J
Jan Kara 已提交
2518 2519 2520
			/* Release allocated io_end */
			ext4_put_io_end(mpd.io_submit.io_end);
			break;
2521
		}
2522

J
Jan Kara 已提交
2523 2524 2525 2526
		trace_ext4_da_write_pages(inode, mpd.first_page, mpd.wbc);
		ret = mpage_prepare_extent_to_map(&mpd);
		if (!ret) {
			if (mpd.map.m_len)
2527 2528
				ret = mpage_map_and_submit_extent(handle, &mpd,
					&give_up_on_write);
J
Jan Kara 已提交
2529 2530 2531 2532 2533 2534 2535 2536 2537
			else {
				/*
				 * We scanned the whole range (or exhausted
				 * nr_to_write), submitted what was mapped and
				 * didn't find anything needing mapping. We are
				 * done.
				 */
				done = true;
			}
2538
		}
2539
		ext4_journal_stop(handle);
J
Jan Kara 已提交
2540 2541 2542
		/* Submit prepared bio */
		ext4_io_submit(&mpd.io_submit);
		/* Unlock pages we didn't use */
2543
		mpage_release_unused_pages(&mpd, give_up_on_write);
J
Jan Kara 已提交
2544 2545 2546 2547 2548 2549
		/* Drop our io_end reference we got from init */
		ext4_put_io_end(mpd.io_submit.io_end);

		if (ret == -ENOSPC && sbi->s_journal) {
			/*
			 * Commit the transaction which would
2550 2551 2552
			 * free blocks released in the transaction
			 * and try again
			 */
2553
			jbd2_journal_force_commit_nested(sbi->s_journal);
2554
			ret = 0;
J
Jan Kara 已提交
2555 2556 2557 2558
			continue;
		}
		/* Fatal error - ENOMEM, EIO... */
		if (ret)
2559
			break;
2560
	}
S
Shaohua Li 已提交
2561
	blk_finish_plug(&plug);
J
Jan Kara 已提交
2562
	if (!ret && !cycled) {
2563
		cycled = 1;
J
Jan Kara 已提交
2564 2565
		mpd.last_page = writeback_index - 1;
		mpd.first_page = 0;
2566 2567
		goto retry;
	}
2568 2569 2570 2571

	/* Update index */
	if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
		/*
J
Jan Kara 已提交
2572
		 * Set the writeback_index so that range_cyclic
2573 2574
		 * mode will write it back later
		 */
J
Jan Kara 已提交
2575
		mapping->writeback_index = mpd.first_page;
2576

2577
out_writepages:
2578 2579
	trace_ext4_writepages_result(inode, wbc, ret,
				     nr_to_write - wbc->nr_to_write);
2580
	return ret;
2581 2582
}

2583 2584
static int ext4_nonda_switch(struct super_block *sb)
{
2585
	s64 free_clusters, dirty_clusters;
2586 2587 2588 2589 2590
	struct ext4_sb_info *sbi = EXT4_SB(sb);

	/*
	 * switch to non delalloc mode if we are running low
	 * on free block. The free block accounting via percpu
2591
	 * counters can get slightly wrong with percpu_counter_batch getting
2592 2593 2594 2595
	 * accumulated on each CPU without updating global counters
	 * Delalloc need an accurate free block accounting. So switch
	 * to non delalloc when we are near to error range.
	 */
2596 2597 2598 2599
	free_clusters =
		percpu_counter_read_positive(&sbi->s_freeclusters_counter);
	dirty_clusters =
		percpu_counter_read_positive(&sbi->s_dirtyclusters_counter);
2600 2601 2602
	/*
	 * Start pushing delalloc when 1/2 of free blocks are dirty.
	 */
2603
	if (dirty_clusters && (free_clusters < 2 * dirty_clusters))
2604
		try_to_writeback_inodes_sb(sb, WB_REASON_FS_FREE_SPACE);
2605

2606 2607
	if (2 * free_clusters < 3 * dirty_clusters ||
	    free_clusters < (dirty_clusters + EXT4_FREECLUSTERS_WATERMARK)) {
2608
		/*
2609 2610
		 * free block count is less than 150% of dirty blocks
		 * or free blocks is less than watermark
2611 2612 2613 2614 2615 2616
		 */
		return 1;
	}
	return 0;
}

2617
static int ext4_da_write_begin(struct file *file, struct address_space *mapping,
2618 2619
			       loff_t pos, unsigned len, unsigned flags,
			       struct page **pagep, void **fsdata)
2620
{
2621
	int ret, retries = 0;
2622 2623 2624 2625 2626 2627
	struct page *page;
	pgoff_t index;
	struct inode *inode = mapping->host;
	handle_t *handle;

	index = pos >> PAGE_CACHE_SHIFT;
2628 2629 2630 2631 2632 2633 2634

	if (ext4_nonda_switch(inode->i_sb)) {
		*fsdata = (void *)FALL_BACK_TO_NONDELALLOC;
		return ext4_write_begin(file, mapping, pos,
					len, flags, pagep, fsdata);
	}
	*fsdata = (void *)0;
2635
	trace_ext4_da_write_begin(inode, pos, len, flags);
2636 2637 2638 2639 2640 2641

	if (ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA)) {
		ret = ext4_da_write_inline_data_begin(mapping, inode,
						      pos, len, flags,
						      pagep, fsdata);
		if (ret < 0)
2642 2643 2644
			return ret;
		if (ret == 1)
			return 0;
2645 2646
	}

2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659
	/*
	 * grab_cache_page_write_begin() can take a long time if the
	 * system is thrashing due to memory pressure, or if the page
	 * is being written back.  So grab it first before we start
	 * the transaction handle.  This also allows us to allocate
	 * the page (if needed) without using GFP_NOFS.
	 */
retry_grab:
	page = grab_cache_page_write_begin(mapping, index, flags);
	if (!page)
		return -ENOMEM;
	unlock_page(page);

2660 2661 2662 2663 2664 2665
	/*
	 * With delayed allocation, we don't log the i_disksize update
	 * if there is delayed block allocation. But we still need
	 * to journalling the i_disksize update if writes to the end
	 * of file which has an already mapped buffer.
	 */
2666
retry_journal:
2667
	handle = ext4_journal_start(inode, EXT4_HT_WRITE_PAGE, 1);
2668
	if (IS_ERR(handle)) {
2669 2670
		page_cache_release(page);
		return PTR_ERR(handle);
2671 2672
	}

2673 2674 2675 2676 2677
	lock_page(page);
	if (page->mapping != mapping) {
		/* The page got truncated from under us */
		unlock_page(page);
		page_cache_release(page);
2678
		ext4_journal_stop(handle);
2679
		goto retry_grab;
2680
	}
2681
	/* In case writeback began while the page was unlocked */
2682
	wait_for_stable_page(page);
2683

2684
	ret = __block_write_begin(page, pos, len, ext4_da_get_block_prep);
2685 2686 2687
	if (ret < 0) {
		unlock_page(page);
		ext4_journal_stop(handle);
2688 2689 2690 2691 2692 2693
		/*
		 * block_write_begin may have instantiated a few blocks
		 * outside i_size.  Trim these off again. Don't need
		 * i_size_read because we hold i_mutex.
		 */
		if (pos + len > inode->i_size)
2694
			ext4_truncate_failed_write(inode);
2695 2696 2697 2698 2699 2700 2701

		if (ret == -ENOSPC &&
		    ext4_should_retry_alloc(inode->i_sb, &retries))
			goto retry_journal;

		page_cache_release(page);
		return ret;
2702 2703
	}

2704
	*pagep = page;
2705 2706 2707
	return ret;
}

2708 2709 2710 2711 2712
/*
 * Check if we should update i_disksize
 * when write to the end of file but not require block allocation
 */
static int ext4_da_should_update_i_disksize(struct page *page,
2713
					    unsigned long offset)
2714 2715 2716 2717 2718 2719 2720 2721 2722
{
	struct buffer_head *bh;
	struct inode *inode = page->mapping->host;
	unsigned int idx;
	int i;

	bh = page_buffers(page);
	idx = offset >> inode->i_blkbits;

2723
	for (i = 0; i < idx; i++)
2724 2725
		bh = bh->b_this_page;

2726
	if (!buffer_mapped(bh) || (buffer_delay(bh)) || buffer_unwritten(bh))
2727 2728 2729 2730
		return 0;
	return 1;
}

2731
static int ext4_da_write_end(struct file *file,
2732 2733 2734
			     struct address_space *mapping,
			     loff_t pos, unsigned len, unsigned copied,
			     struct page *page, void *fsdata)
2735 2736 2737 2738 2739
{
	struct inode *inode = mapping->host;
	int ret = 0, ret2;
	handle_t *handle = ext4_journal_current_handle();
	loff_t new_i_size;
2740
	unsigned long start, end;
2741 2742
	int write_mode = (int)(unsigned long)fsdata;

2743 2744 2745
	if (write_mode == FALL_BACK_TO_NONDELALLOC)
		return ext4_write_end(file, mapping, pos,
				      len, copied, page, fsdata);
2746

2747
	trace_ext4_da_write_end(inode, pos, len, copied);
2748
	start = pos & (PAGE_CACHE_SIZE - 1);
2749
	end = start + copied - 1;
2750 2751 2752 2753 2754 2755 2756

	/*
	 * generic_write_end() will run mark_inode_dirty() if i_size
	 * changes.  So let's piggyback the i_disksize mark_inode_dirty
	 * into that.
	 */
	new_i_size = pos + copied;
2757
	if (copied && new_i_size > EXT4_I(inode)->i_disksize) {
2758 2759
		if (ext4_has_inline_data(inode) ||
		    ext4_da_should_update_i_disksize(page, end)) {
2760
			down_write(&EXT4_I(inode)->i_data_sem);
2761
			if (new_i_size > EXT4_I(inode)->i_disksize)
2762 2763
				EXT4_I(inode)->i_disksize = new_i_size;
			up_write(&EXT4_I(inode)->i_data_sem);
2764 2765 2766 2767 2768
			/* We need to mark inode dirty even if
			 * new_i_size is less that inode->i_size
			 * bu greater than i_disksize.(hint delalloc)
			 */
			ext4_mark_inode_dirty(handle, inode);
2769
		}
2770
	}
2771 2772 2773 2774 2775 2776 2777 2778

	if (write_mode != CONVERT_INLINE_DATA &&
	    ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA) &&
	    ext4_has_inline_data(inode))
		ret2 = ext4_da_write_inline_data_end(inode, pos, len, copied,
						     page);
	else
		ret2 = generic_write_end(file, mapping, pos, len, copied,
2779
							page, fsdata);
2780

2781 2782 2783 2784 2785 2786 2787 2788 2789 2790
	copied = ret2;
	if (ret2 < 0)
		ret = ret2;
	ret2 = ext4_journal_stop(handle);
	if (!ret)
		ret = ret2;

	return ret ? ret : copied;
}

2791 2792
static void ext4_da_invalidatepage(struct page *page, unsigned int offset,
				   unsigned int length)
2793 2794 2795 2796 2797 2798 2799 2800
{
	/*
	 * Drop reserved blocks
	 */
	BUG_ON(!PageLocked(page));
	if (!page_has_buffers(page))
		goto out;

2801
	ext4_da_page_release_reservation(page, offset, length);
2802 2803

out:
2804
	ext4_invalidatepage(page, offset, length);
2805 2806 2807 2808

	return;
}

2809 2810 2811 2812 2813
/*
 * Force all delayed allocation blocks to be allocated for a given inode.
 */
int ext4_alloc_da_blocks(struct inode *inode)
{
2814 2815
	trace_ext4_alloc_da_blocks(inode);

2816 2817 2818 2819 2820 2821 2822 2823 2824 2825
	if (!EXT4_I(inode)->i_reserved_data_blocks &&
	    !EXT4_I(inode)->i_reserved_meta_blocks)
		return 0;

	/*
	 * We do something simple for now.  The filemap_flush() will
	 * also start triggering a write of the data blocks, which is
	 * not strictly speaking necessary (and for users of
	 * laptop_mode, not even desirable).  However, to do otherwise
	 * would require replicating code paths in:
2826
	 *
2827
	 * ext4_writepages() ->
2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838
	 *    write_cache_pages() ---> (via passed in callback function)
	 *        __mpage_da_writepage() -->
	 *           mpage_add_bh_to_extent()
	 *           mpage_da_map_blocks()
	 *
	 * The problem is that write_cache_pages(), located in
	 * mm/page-writeback.c, marks pages clean in preparation for
	 * doing I/O, which is not desirable if we're not planning on
	 * doing I/O at all.
	 *
	 * We could call write_cache_pages(), and then redirty all of
2839
	 * the pages by calling redirty_page_for_writepage() but that
2840 2841
	 * would be ugly in the extreme.  So instead we would need to
	 * replicate parts of the code in the above functions,
L
Lucas De Marchi 已提交
2842
	 * simplifying them because we wouldn't actually intend to
2843 2844 2845
	 * write out the pages, but rather only collect contiguous
	 * logical block extents, call the multi-block allocator, and
	 * then update the buffer heads with the block allocations.
2846
	 *
2847 2848 2849 2850 2851 2852
	 * For now, though, we'll cheat by calling filemap_flush(),
	 * which will map the blocks, and start the I/O, but not
	 * actually wait for the I/O to complete.
	 */
	return filemap_flush(inode->i_mapping);
}
2853

2854 2855 2856 2857 2858
/*
 * bmap() is special.  It gets used by applications such as lilo and by
 * the swapper to find the on-disk block of a specific piece of data.
 *
 * Naturally, this is dangerous if the block concerned is still in the
2859
 * journal.  If somebody makes a swapfile on an ext4 data-journaling
2860 2861 2862 2863 2864 2865 2866 2867
 * filesystem and enables swap, then they may get a nasty shock when the
 * data getting swapped to that swapfile suddenly gets overwritten by
 * the original zero's written out previously to the journal and
 * awaiting writeback in the kernel's buffer cache.
 *
 * So, if we see any bmap calls here on a modified, data-journaled file,
 * take extra steps to flush any blocks which might be in the cache.
 */
2868
static sector_t ext4_bmap(struct address_space *mapping, sector_t block)
2869 2870 2871 2872 2873
{
	struct inode *inode = mapping->host;
	journal_t *journal;
	int err;

T
Tao Ma 已提交
2874 2875 2876 2877 2878 2879
	/*
	 * We can get here for an inline file via the FIBMAP ioctl
	 */
	if (ext4_has_inline_data(inode))
		return 0;

2880 2881 2882 2883 2884 2885 2886 2887 2888 2889
	if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) &&
			test_opt(inode->i_sb, DELALLOC)) {
		/*
		 * With delalloc we want to sync the file
		 * so that we can make sure we allocate
		 * blocks for file
		 */
		filemap_write_and_wait(mapping);
	}

2890 2891
	if (EXT4_JOURNAL(inode) &&
	    ext4_test_inode_state(inode, EXT4_STATE_JDATA)) {
2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902
		/*
		 * This is a REALLY heavyweight approach, but the use of
		 * bmap on dirty files is expected to be extremely rare:
		 * only if we run lilo or swapon on a freshly made file
		 * do we expect this to happen.
		 *
		 * (bmap requires CAP_SYS_RAWIO so this does not
		 * represent an unprivileged user DOS attack --- we'd be
		 * in trouble if mortal users could trigger this path at
		 * will.)
		 *
2903
		 * NB. EXT4_STATE_JDATA is not set on files other than
2904 2905 2906 2907 2908 2909
		 * regular files.  If somebody wants to bmap a directory
		 * or symlink and gets confused because the buffer
		 * hasn't yet been flushed to disk, they deserve
		 * everything they get.
		 */

2910
		ext4_clear_inode_state(inode, EXT4_STATE_JDATA);
2911
		journal = EXT4_JOURNAL(inode);
2912 2913 2914
		jbd2_journal_lock_updates(journal);
		err = jbd2_journal_flush(journal);
		jbd2_journal_unlock_updates(journal);
2915 2916 2917 2918 2919

		if (err)
			return 0;
	}

2920
	return generic_block_bmap(mapping, block, ext4_get_block);
2921 2922
}

2923
static int ext4_readpage(struct file *file, struct page *page)
2924
{
T
Tao Ma 已提交
2925 2926 2927
	int ret = -EAGAIN;
	struct inode *inode = page->mapping->host;

2928
	trace_ext4_readpage(page);
T
Tao Ma 已提交
2929 2930 2931 2932 2933 2934 2935 2936

	if (ext4_has_inline_data(inode))
		ret = ext4_readpage_inline(inode, page);

	if (ret == -EAGAIN)
		return mpage_readpage(page, ext4_get_block);

	return ret;
2937 2938 2939
}

static int
2940
ext4_readpages(struct file *file, struct address_space *mapping,
2941 2942
		struct list_head *pages, unsigned nr_pages)
{
T
Tao Ma 已提交
2943 2944 2945 2946 2947 2948
	struct inode *inode = mapping->host;

	/* If the file has inline data, no need to do readpages. */
	if (ext4_has_inline_data(inode))
		return 0;

2949
	return mpage_readpages(mapping, pages, nr_pages, ext4_get_block);
2950 2951
}

2952 2953
static void ext4_invalidatepage(struct page *page, unsigned int offset,
				unsigned int length)
2954
{
2955
	trace_ext4_invalidatepage(page, offset, length);
2956

2957 2958 2959
	/* No journalling happens on data buffers when this function is used */
	WARN_ON(page_has_buffers(page) && buffer_jbd(page_buffers(page)));

2960
	block_invalidatepage(page, offset, length);
2961 2962
}

2963
static int __ext4_journalled_invalidatepage(struct page *page,
2964 2965
					    unsigned int offset,
					    unsigned int length)
2966 2967 2968
{
	journal_t *journal = EXT4_JOURNAL(page->mapping->host);

2969
	trace_ext4_journalled_invalidatepage(page, offset, length);
2970

2971 2972 2973
	/*
	 * If it's a full truncate we just forget about the pending dirtying
	 */
2974
	if (offset == 0 && length == PAGE_CACHE_SIZE)
2975 2976
		ClearPageChecked(page);

2977
	return jbd2_journal_invalidatepage(journal, page, offset, length);
2978 2979 2980 2981
}

/* Wrapper for aops... */
static void ext4_journalled_invalidatepage(struct page *page,
2982 2983
					   unsigned int offset,
					   unsigned int length)
2984
{
2985
	WARN_ON(__ext4_journalled_invalidatepage(page, offset, length) < 0);
2986 2987
}

2988
static int ext4_releasepage(struct page *page, gfp_t wait)
2989
{
2990
	journal_t *journal = EXT4_JOURNAL(page->mapping->host);
2991

2992 2993
	trace_ext4_releasepage(page);

2994 2995
	/* Page has dirty journalled data -> cannot release */
	if (PageChecked(page))
2996
		return 0;
2997 2998 2999 3000
	if (journal)
		return jbd2_journal_try_to_free_buffers(journal, page, wait);
	else
		return try_to_free_buffers(page);
3001 3002
}

3003 3004 3005 3006 3007
/*
 * ext4_get_block used when preparing for a DIO write or buffer write.
 * We allocate an uinitialized extent if blocks haven't been allocated.
 * The extent will be converted to initialized after the IO is complete.
 */
3008
int ext4_get_block_write(struct inode *inode, sector_t iblock,
3009 3010
		   struct buffer_head *bh_result, int create)
{
3011
	ext4_debug("ext4_get_block_write: inode %lu, create flag %d\n",
3012
		   inode->i_ino, create);
3013 3014
	return _ext4_get_block(inode, iblock, bh_result,
			       EXT4_GET_BLOCKS_IO_CREATE_EXT);
3015 3016
}

3017
static int ext4_get_block_write_nolock(struct inode *inode, sector_t iblock,
3018
		   struct buffer_head *bh_result, int create)
3019
{
3020 3021 3022 3023
	ext4_debug("ext4_get_block_write_nolock: inode %lu, create flag %d\n",
		   inode->i_ino, create);
	return _ext4_get_block(inode, iblock, bh_result,
			       EXT4_GET_BLOCKS_NO_LOCK);
3024 3025
}

3026
static void ext4_end_io_dio(struct kiocb *iocb, loff_t offset,
3027 3028
			    ssize_t size, void *private, int ret,
			    bool is_async)
3029
{
A
Al Viro 已提交
3030
	struct inode *inode = file_inode(iocb->ki_filp);
3031 3032
        ext4_io_end_t *io_end = iocb->private;

J
Jan Kara 已提交
3033 3034 3035 3036 3037 3038 3039
	/* if not async direct IO just return */
	if (!io_end) {
		inode_dio_done(inode);
		if (is_async)
			aio_complete(iocb, ret, 0);
		return;
	}
3040

3041
	ext_debug("ext4_end_io_dio(): io_end 0x%p "
3042
		  "for inode %lu, iocb 0x%p, offset %llu, size %zd\n",
3043 3044 3045
 		  iocb->private, io_end->inode->i_ino, iocb, offset,
		  size);

3046
	iocb->private = NULL;
3047 3048
	io_end->offset = offset;
	io_end->size = size;
3049 3050 3051 3052
	if (is_async) {
		io_end->iocb = iocb;
		io_end->result = ret;
	}
J
Jan Kara 已提交
3053
	ext4_put_io_end_defer(io_end);
3054
}
3055

3056 3057 3058 3059 3060
/*
 * For ext4 extent files, ext4 will do direct-io write to holes,
 * preallocated extents, and those write extend the file, no need to
 * fall back to buffered IO.
 *
3061
 * For holes, we fallocate those blocks, mark them as uninitialized
3062
 * If those blocks were preallocated, we mark sure they are split, but
3063
 * still keep the range to write as uninitialized.
3064
 *
3065
 * The unwritten extents will be converted to written when DIO is completed.
3066
 * For async direct IO, since the IO may still pending when return, we
L
Lucas De Marchi 已提交
3067
 * set up an end_io call back function, which will do the conversion
3068
 * when async direct IO completed.
3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082
 *
 * If the O_DIRECT write will extend the file then add this inode to the
 * orphan list.  So recovery will truncate it back to the original size
 * if the machine crashes during the write.
 *
 */
static ssize_t ext4_ext_direct_IO(int rw, struct kiocb *iocb,
			      const struct iovec *iov, loff_t offset,
			      unsigned long nr_segs)
{
	struct file *file = iocb->ki_filp;
	struct inode *inode = file->f_mapping->host;
	ssize_t ret;
	size_t count = iov_length(iov, nr_segs);
3083 3084 3085
	int overwrite = 0;
	get_block_t *get_block_func = NULL;
	int dio_flags = 0;
3086
	loff_t final_size = offset + count;
J
Jan Kara 已提交
3087
	ext4_io_end_t *io_end = NULL;
3088

3089 3090 3091
	/* Use the old path for reads and writes beyond i_size. */
	if (rw != WRITE || final_size > inode->i_size)
		return ext4_ind_direct_IO(rw, iocb, iov, offset, nr_segs);
3092

3093
	BUG_ON(iocb->private == NULL);
3094

3095 3096 3097 3098 3099 3100 3101 3102
	/*
	 * Make all waiters for direct IO properly wait also for extent
	 * conversion. This also disallows race between truncate() and
	 * overwrite DIO as i_dio_count needs to be incremented under i_mutex.
	 */
	if (rw == WRITE)
		atomic_inc(&inode->i_dio_count);

3103 3104
	/* If we do a overwrite dio, i_mutex locking can be released */
	overwrite = *((int *)iocb->private);
3105

3106 3107 3108 3109
	if (overwrite) {
		down_read(&EXT4_I(inode)->i_data_sem);
		mutex_unlock(&inode->i_mutex);
	}
3110

3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132
	/*
	 * We could direct write to holes and fallocate.
	 *
	 * Allocated blocks to fill the hole are marked as
	 * uninitialized to prevent parallel buffered read to expose
	 * the stale data before DIO complete the data IO.
	 *
	 * As to previously fallocated extents, ext4 get_block will
	 * just simply mark the buffer mapped but still keep the
	 * extents uninitialized.
	 *
	 * For non AIO case, we will convert those unwritten extents
	 * to written after return back from blockdev_direct_IO.
	 *
	 * For async DIO, the conversion needs to be deferred when the
	 * IO is completed. The ext4 end_io callback function will be
	 * called to take care of the conversion work.  Here for async
	 * case, we allocate an io_end structure to hook to the iocb.
	 */
	iocb->private = NULL;
	ext4_inode_aio_set(inode, NULL);
	if (!is_sync_kiocb(iocb)) {
J
Jan Kara 已提交
3133
		io_end = ext4_init_io_end(inode, GFP_NOFS);
3134 3135 3136
		if (!io_end) {
			ret = -ENOMEM;
			goto retake_lock;
3137
		}
3138
		io_end->flag |= EXT4_IO_END_DIRECT;
J
Jan Kara 已提交
3139 3140 3141 3142
		/*
		 * Grab reference for DIO. Will be dropped in ext4_end_io_dio()
		 */
		iocb->private = ext4_get_io_end(io_end);
3143
		/*
3144 3145 3146 3147
		 * we save the io structure for current async direct
		 * IO, so that later ext4_map_blocks() could flag the
		 * io structure whether there is a unwritten extents
		 * needs to be converted when IO is completed.
3148
		 */
3149 3150
		ext4_inode_aio_set(inode, io_end);
	}
3151

3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166
	if (overwrite) {
		get_block_func = ext4_get_block_write_nolock;
	} else {
		get_block_func = ext4_get_block_write;
		dio_flags = DIO_LOCKING;
	}
	ret = __blockdev_direct_IO(rw, iocb, inode,
				   inode->i_sb->s_bdev, iov,
				   offset, nr_segs,
				   get_block_func,
				   ext4_end_io_dio,
				   NULL,
				   dio_flags);

	/*
J
Jan Kara 已提交
3167 3168 3169 3170 3171
	 * Put our reference to io_end. This can free the io_end structure e.g.
	 * in sync IO case or in case of error. It can even perform extent
	 * conversion if all bios we submitted finished before we got here.
	 * Note that in that case iocb->private can be already set to NULL
	 * here.
3172
	 */
J
Jan Kara 已提交
3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194
	if (io_end) {
		ext4_inode_aio_set(inode, NULL);
		ext4_put_io_end(io_end);
		/*
		 * When no IO was submitted ext4_end_io_dio() was not
		 * called so we have to put iocb's reference.
		 */
		if (ret <= 0 && ret != -EIOCBQUEUED && iocb->private) {
			WARN_ON(iocb->private != io_end);
			WARN_ON(io_end->flag & EXT4_IO_END_UNWRITTEN);
			WARN_ON(io_end->iocb);
			/*
			 * Generic code already did inode_dio_done() so we
			 * have to clear EXT4_IO_END_DIRECT to not do it for
			 * the second time.
			 */
			io_end->flag = 0;
			ext4_put_io_end(io_end);
			iocb->private = NULL;
		}
	}
	if (ret > 0 && !overwrite && ext4_test_inode_state(inode,
3195 3196 3197 3198 3199 3200
						EXT4_STATE_DIO_UNWRITTEN)) {
		int err;
		/*
		 * for non AIO case, since the IO is already
		 * completed, we could do the conversion right here
		 */
3201
		err = ext4_convert_unwritten_extents(NULL, inode,
3202 3203 3204 3205 3206
						     offset, ret);
		if (err < 0)
			ret = err;
		ext4_clear_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN);
	}
3207

3208
retake_lock:
3209 3210
	if (rw == WRITE)
		inode_dio_done(inode);
3211 3212 3213 3214
	/* take i_mutex locking again if we do a ovewrite dio */
	if (overwrite) {
		up_read(&EXT4_I(inode)->i_data_sem);
		mutex_lock(&inode->i_mutex);
3215
	}
3216

3217
	return ret;
3218 3219 3220 3221 3222 3223 3224 3225
}

static ssize_t ext4_direct_IO(int rw, struct kiocb *iocb,
			      const struct iovec *iov, loff_t offset,
			      unsigned long nr_segs)
{
	struct file *file = iocb->ki_filp;
	struct inode *inode = file->f_mapping->host;
3226
	ssize_t ret;
3227

3228 3229 3230 3231 3232 3233
	/*
	 * If we are doing data journalling we don't support O_DIRECT
	 */
	if (ext4_should_journal_data(inode))
		return 0;

T
Tao Ma 已提交
3234 3235 3236 3237
	/* Let buffer I/O handle the inline data case. */
	if (ext4_has_inline_data(inode))
		return 0;

3238
	trace_ext4_direct_IO_enter(inode, offset, iov_length(iov, nr_segs), rw);
3239
	if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
3240 3241 3242 3243 3244 3245
		ret = ext4_ext_direct_IO(rw, iocb, iov, offset, nr_segs);
	else
		ret = ext4_ind_direct_IO(rw, iocb, iov, offset, nr_segs);
	trace_ext4_direct_IO_exit(inode, offset,
				iov_length(iov, nr_segs), rw, ret);
	return ret;
3246 3247
}

3248
/*
3249
 * Pages can be marked dirty completely asynchronously from ext4's journalling
3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260
 * activity.  By filemap_sync_pte(), try_to_unmap_one(), etc.  We cannot do
 * much here because ->set_page_dirty is called under VFS locks.  The page is
 * not necessarily locked.
 *
 * We cannot just dirty the page and leave attached buffers clean, because the
 * buffers' dirty state is "definitive".  We cannot just set the buffers dirty
 * or jbddirty because all the journalling code will explode.
 *
 * So what we do is to mark the page "pending dirty" and next time writepage
 * is called, propagate that into the buffers appropriately.
 */
3261
static int ext4_journalled_set_page_dirty(struct page *page)
3262 3263 3264 3265 3266
{
	SetPageChecked(page);
	return __set_page_dirty_nobuffers(page);
}

3267
static const struct address_space_operations ext4_aops = {
3268 3269
	.readpage		= ext4_readpage,
	.readpages		= ext4_readpages,
3270
	.writepage		= ext4_writepage,
3271
	.writepages		= ext4_writepages,
3272
	.write_begin		= ext4_write_begin,
3273
	.write_end		= ext4_write_end,
3274 3275 3276 3277 3278 3279
	.bmap			= ext4_bmap,
	.invalidatepage		= ext4_invalidatepage,
	.releasepage		= ext4_releasepage,
	.direct_IO		= ext4_direct_IO,
	.migratepage		= buffer_migrate_page,
	.is_partially_uptodate  = block_is_partially_uptodate,
3280
	.error_remove_page	= generic_error_remove_page,
3281 3282
};

3283
static const struct address_space_operations ext4_journalled_aops = {
3284 3285
	.readpage		= ext4_readpage,
	.readpages		= ext4_readpages,
3286
	.writepage		= ext4_writepage,
3287
	.writepages		= ext4_writepages,
3288 3289 3290 3291
	.write_begin		= ext4_write_begin,
	.write_end		= ext4_journalled_write_end,
	.set_page_dirty		= ext4_journalled_set_page_dirty,
	.bmap			= ext4_bmap,
3292
	.invalidatepage		= ext4_journalled_invalidatepage,
3293
	.releasepage		= ext4_releasepage,
3294
	.direct_IO		= ext4_direct_IO,
3295
	.is_partially_uptodate  = block_is_partially_uptodate,
3296
	.error_remove_page	= generic_error_remove_page,
3297 3298
};

3299
static const struct address_space_operations ext4_da_aops = {
3300 3301
	.readpage		= ext4_readpage,
	.readpages		= ext4_readpages,
3302
	.writepage		= ext4_writepage,
3303
	.writepages		= ext4_writepages,
3304 3305 3306 3307 3308 3309 3310 3311
	.write_begin		= ext4_da_write_begin,
	.write_end		= ext4_da_write_end,
	.bmap			= ext4_bmap,
	.invalidatepage		= ext4_da_invalidatepage,
	.releasepage		= ext4_releasepage,
	.direct_IO		= ext4_direct_IO,
	.migratepage		= buffer_migrate_page,
	.is_partially_uptodate  = block_is_partially_uptodate,
3312
	.error_remove_page	= generic_error_remove_page,
3313 3314
};

3315
void ext4_set_aops(struct inode *inode)
3316
{
3317 3318
	switch (ext4_inode_journal_mode(inode)) {
	case EXT4_INODE_ORDERED_DATA_MODE:
3319
		ext4_set_inode_state(inode, EXT4_STATE_ORDERED_MODE);
3320 3321
		break;
	case EXT4_INODE_WRITEBACK_DATA_MODE:
3322
		ext4_clear_inode_state(inode, EXT4_STATE_ORDERED_MODE);
3323 3324
		break;
	case EXT4_INODE_JOURNAL_DATA_MODE:
3325
		inode->i_mapping->a_ops = &ext4_journalled_aops;
3326
		return;
3327 3328 3329
	default:
		BUG();
	}
3330 3331 3332 3333
	if (test_opt(inode->i_sb, DELALLOC))
		inode->i_mapping->a_ops = &ext4_da_aops;
	else
		inode->i_mapping->a_ops = &ext4_aops;
3334 3335
}

3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439
/*
 * ext4_block_truncate_page() zeroes out a mapping from file offset `from'
 * up to the end of the block which corresponds to `from'.
 * This required during truncate. We need to physically zero the tail end
 * of that block so it doesn't yield old data if the file is later grown.
 */
int ext4_block_truncate_page(handle_t *handle,
		struct address_space *mapping, loff_t from)
{
	unsigned offset = from & (PAGE_CACHE_SIZE-1);
	unsigned length;
	unsigned blocksize;
	struct inode *inode = mapping->host;

	blocksize = inode->i_sb->s_blocksize;
	length = blocksize - (offset & (blocksize - 1));

	return ext4_block_zero_page_range(handle, mapping, from, length);
}

/*
 * ext4_block_zero_page_range() zeros out a mapping of length 'length'
 * starting from file offset 'from'.  The range to be zero'd must
 * be contained with in one block.  If the specified range exceeds
 * the end of the block it will be shortened to end of the block
 * that cooresponds to 'from'
 */
int ext4_block_zero_page_range(handle_t *handle,
		struct address_space *mapping, loff_t from, loff_t length)
{
	ext4_fsblk_t index = from >> PAGE_CACHE_SHIFT;
	unsigned offset = from & (PAGE_CACHE_SIZE-1);
	unsigned blocksize, max, pos;
	ext4_lblk_t iblock;
	struct inode *inode = mapping->host;
	struct buffer_head *bh;
	struct page *page;
	int err = 0;

	page = find_or_create_page(mapping, from >> PAGE_CACHE_SHIFT,
				   mapping_gfp_mask(mapping) & ~__GFP_FS);
	if (!page)
		return -ENOMEM;

	blocksize = inode->i_sb->s_blocksize;
	max = blocksize - (offset & (blocksize - 1));

	/*
	 * correct length if it does not fall between
	 * 'from' and the end of the block
	 */
	if (length > max || length < 0)
		length = max;

	iblock = index << (PAGE_CACHE_SHIFT - inode->i_sb->s_blocksize_bits);

	if (!page_has_buffers(page))
		create_empty_buffers(page, blocksize, 0);

	/* Find the buffer that contains "offset" */
	bh = page_buffers(page);
	pos = blocksize;
	while (offset >= pos) {
		bh = bh->b_this_page;
		iblock++;
		pos += blocksize;
	}
	if (buffer_freed(bh)) {
		BUFFER_TRACE(bh, "freed: skip");
		goto unlock;
	}
	if (!buffer_mapped(bh)) {
		BUFFER_TRACE(bh, "unmapped");
		ext4_get_block(inode, iblock, bh, 0);
		/* unmapped? It's a hole - nothing to do */
		if (!buffer_mapped(bh)) {
			BUFFER_TRACE(bh, "still unmapped");
			goto unlock;
		}
	}

	/* Ok, it's mapped. Make sure it's up-to-date */
	if (PageUptodate(page))
		set_buffer_uptodate(bh);

	if (!buffer_uptodate(bh)) {
		err = -EIO;
		ll_rw_block(READ, 1, &bh);
		wait_on_buffer(bh);
		/* Uhhuh. Read error. Complain and punt. */
		if (!buffer_uptodate(bh))
			goto unlock;
	}
	if (ext4_should_journal_data(inode)) {
		BUFFER_TRACE(bh, "get write access");
		err = ext4_journal_get_write_access(handle, bh);
		if (err)
			goto unlock;
	}
	zero_user(page, offset, length);
	BUFFER_TRACE(bh, "zeroed end of block");

	if (ext4_should_journal_data(inode)) {
		err = ext4_handle_dirty_metadata(handle, inode, bh);
3440
	} else {
3441
		err = 0;
3442
		mark_buffer_dirty(bh);
3443 3444 3445
		if (ext4_test_inode_state(inode, EXT4_STATE_ORDERED_MODE))
			err = ext4_jbd2_file_inode(handle, inode);
	}
3446 3447 3448 3449 3450 3451 3452

unlock:
	unlock_page(page);
	page_cache_release(page);
	return err;
}

3453 3454 3455 3456 3457
int ext4_zero_partial_blocks(handle_t *handle, struct inode *inode,
			     loff_t lstart, loff_t length)
{
	struct super_block *sb = inode->i_sb;
	struct address_space *mapping = inode->i_mapping;
3458
	unsigned partial_start, partial_end;
3459 3460 3461 3462
	ext4_fsblk_t start, end;
	loff_t byte_end = (lstart + length - 1);
	int err = 0;

3463 3464 3465
	partial_start = lstart & (sb->s_blocksize - 1);
	partial_end = byte_end & (sb->s_blocksize - 1);

3466 3467 3468 3469
	start = lstart >> sb->s_blocksize_bits;
	end = byte_end >> sb->s_blocksize_bits;

	/* Handle partial zero within the single block */
3470 3471
	if (start == end &&
	    (partial_start || (partial_end != sb->s_blocksize - 1))) {
3472 3473 3474 3475 3476
		err = ext4_block_zero_page_range(handle, mapping,
						 lstart, length);
		return err;
	}
	/* Handle partial zero out on the start of the range */
3477
	if (partial_start) {
3478 3479 3480 3481 3482 3483
		err = ext4_block_zero_page_range(handle, mapping,
						 lstart, sb->s_blocksize);
		if (err)
			return err;
	}
	/* Handle partial zero out on the end of the range */
3484
	if (partial_end != sb->s_blocksize - 1)
3485
		err = ext4_block_zero_page_range(handle, mapping,
3486 3487
						 byte_end - partial_end,
						 partial_end + 1);
3488 3489 3490
	return err;
}

3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501
int ext4_can_truncate(struct inode *inode)
{
	if (S_ISREG(inode->i_mode))
		return 1;
	if (S_ISDIR(inode->i_mode))
		return 1;
	if (S_ISLNK(inode->i_mode))
		return !ext4_inode_is_fast_symlink(inode);
	return 0;
}

3502 3503 3504 3505 3506 3507 3508 3509
/*
 * ext4_punch_hole: punches a hole in a file by releaseing the blocks
 * associated with the given offset and length
 *
 * @inode:  File inode
 * @offset: The offset where the hole will begin
 * @len:    The length of the hole
 *
3510
 * Returns: 0 on success or negative on failure
3511 3512
 */

3513
int ext4_punch_hole(struct inode *inode, loff_t offset, loff_t length)
3514
{
T
Theodore Ts'o 已提交
3515 3516 3517
	struct super_block *sb = inode->i_sb;
	ext4_lblk_t first_block, stop_block;
	struct address_space *mapping = inode->i_mapping;
3518
	loff_t first_block_offset, last_block_offset;
T
Theodore Ts'o 已提交
3519 3520 3521 3522
	handle_t *handle;
	unsigned int credits;
	int ret = 0;

3523
	if (!S_ISREG(inode->i_mode))
3524
		return -EOPNOTSUPP;
3525

T
Theodore Ts'o 已提交
3526
	if (EXT4_SB(sb)->s_cluster_ratio > 1) {
3527
		/* TODO: Add support for bigalloc file systems */
3528
		return -EOPNOTSUPP;
3529 3530
	}

3531 3532
	trace_ext4_punch_hole(inode, offset, length);

T
Theodore Ts'o 已提交
3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568
	/*
	 * Write out all dirty pages to avoid race conditions
	 * Then release them.
	 */
	if (mapping->nrpages && mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) {
		ret = filemap_write_and_wait_range(mapping, offset,
						   offset + length - 1);
		if (ret)
			return ret;
	}

	mutex_lock(&inode->i_mutex);
	/* It's not possible punch hole on append only file */
	if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) {
		ret = -EPERM;
		goto out_mutex;
	}
	if (IS_SWAPFILE(inode)) {
		ret = -ETXTBSY;
		goto out_mutex;
	}

	/* No need to punch hole beyond i_size */
	if (offset >= inode->i_size)
		goto out_mutex;

	/*
	 * If the hole extends beyond i_size, set the hole
	 * to end after the page that contains i_size
	 */
	if (offset + length > inode->i_size) {
		length = inode->i_size +
		   PAGE_CACHE_SIZE - (inode->i_size & (PAGE_CACHE_SIZE - 1)) -
		   offset;
	}

3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580
	if (offset & (sb->s_blocksize - 1) ||
	    (offset + length) & (sb->s_blocksize - 1)) {
		/*
		 * Attach jinode to inode for jbd2 if we do any zeroing of
		 * partial block
		 */
		ret = ext4_inode_attach_jinode(inode);
		if (ret < 0)
			goto out_mutex;

	}

3581 3582
	first_block_offset = round_up(offset, sb->s_blocksize);
	last_block_offset = round_down((offset + length), sb->s_blocksize) - 1;
T
Theodore Ts'o 已提交
3583

3584 3585 3586 3587
	/* Now release the pages and zero block aligned part of pages*/
	if (last_block_offset > first_block_offset)
		truncate_pagecache_range(inode, first_block_offset,
					 last_block_offset);
T
Theodore Ts'o 已提交
3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603

	/* Wait all existing dio workers, newcomers will block on i_mutex */
	ext4_inode_block_unlocked_dio(inode);
	inode_dio_wait(inode);

	if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
		credits = ext4_writepage_trans_blocks(inode);
	else
		credits = ext4_blocks_for_truncate(inode);
	handle = ext4_journal_start(inode, EXT4_HT_TRUNCATE, credits);
	if (IS_ERR(handle)) {
		ret = PTR_ERR(handle);
		ext4_std_error(sb, ret);
		goto out_dio;
	}

3604 3605 3606 3607
	ret = ext4_zero_partial_blocks(handle, inode, offset,
				       length);
	if (ret)
		goto out_stop;
T
Theodore Ts'o 已提交
3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634

	first_block = (offset + sb->s_blocksize - 1) >>
		EXT4_BLOCK_SIZE_BITS(sb);
	stop_block = (offset + length) >> EXT4_BLOCK_SIZE_BITS(sb);

	/* If there are no blocks to remove, return now */
	if (first_block >= stop_block)
		goto out_stop;

	down_write(&EXT4_I(inode)->i_data_sem);
	ext4_discard_preallocations(inode);

	ret = ext4_es_remove_extent(inode, first_block,
				    stop_block - first_block);
	if (ret) {
		up_write(&EXT4_I(inode)->i_data_sem);
		goto out_stop;
	}

	if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
		ret = ext4_ext_remove_space(inode, first_block,
					    stop_block - 1);
	else
		ret = ext4_free_hole_blocks(handle, inode, first_block,
					    stop_block);

	ext4_discard_preallocations(inode);
T
Theodore Ts'o 已提交
3635
	up_write(&EXT4_I(inode)->i_data_sem);
T
Theodore Ts'o 已提交
3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646
	if (IS_SYNC(inode))
		ext4_handle_sync(handle);
	inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
	ext4_mark_inode_dirty(handle, inode);
out_stop:
	ext4_journal_stop(handle);
out_dio:
	ext4_inode_resume_unlocked_dio(inode);
out_mutex:
	mutex_unlock(&inode->i_mutex);
	return ret;
3647 3648
}

3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673
int ext4_inode_attach_jinode(struct inode *inode)
{
	struct ext4_inode_info *ei = EXT4_I(inode);
	struct jbd2_inode *jinode;

	if (ei->jinode || !EXT4_SB(inode->i_sb)->s_journal)
		return 0;

	jinode = jbd2_alloc_inode(GFP_KERNEL);
	spin_lock(&inode->i_lock);
	if (!ei->jinode) {
		if (!jinode) {
			spin_unlock(&inode->i_lock);
			return -ENOMEM;
		}
		ei->jinode = jinode;
		jbd2_journal_init_jbd_inode(ei->jinode, inode);
		jinode = NULL;
	}
	spin_unlock(&inode->i_lock);
	if (unlikely(jinode != NULL))
		jbd2_free_inode(jinode);
	return 0;
}

3674
/*
3675
 * ext4_truncate()
3676
 *
3677 3678
 * We block out ext4_get_block() block instantiations across the entire
 * transaction, and VFS/VM ensures that ext4_truncate() cannot run
3679 3680
 * simultaneously on behalf of the same inode.
 *
3681
 * As we work through the truncate and commit bits of it to the journal there
3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694
 * is one core, guiding principle: the file's tree must always be consistent on
 * disk.  We must be able to restart the truncate after a crash.
 *
 * The file's tree may be transiently inconsistent in memory (although it
 * probably isn't), but whenever we close off and commit a journal transaction,
 * the contents of (the filesystem + the journal) must be consistent and
 * restartable.  It's pretty simple, really: bottom up, right to left (although
 * left-to-right works OK too).
 *
 * Note that at recovery time, journal replay occurs *before* the restart of
 * truncate against the orphan inode list.
 *
 * The committed inode has the new, desired i_size (which is the same as
3695
 * i_disksize in this case).  After a crash, ext4_orphan_cleanup() will see
3696
 * that this inode's truncate did not complete and it will again call
3697 3698
 * ext4_truncate() to have another go.  So there will be instantiated blocks
 * to the right of the truncation point in a crashed ext4 filesystem.  But
3699
 * that's fine - as long as they are linked from the inode, the post-crash
3700
 * ext4_truncate() run will find them and release them.
3701
 */
3702
void ext4_truncate(struct inode *inode)
3703
{
T
Theodore Ts'o 已提交
3704 3705 3706 3707 3708
	struct ext4_inode_info *ei = EXT4_I(inode);
	unsigned int credits;
	handle_t *handle;
	struct address_space *mapping = inode->i_mapping;

3709 3710 3711 3712 3713 3714 3715
	/*
	 * There is a possibility that we're either freeing the inode
	 * or it completely new indode. In those cases we might not
	 * have i_mutex locked because it's not necessary.
	 */
	if (!(inode->i_state & (I_NEW|I_FREEING)))
		WARN_ON(!mutex_is_locked(&inode->i_mutex));
3716 3717
	trace_ext4_truncate_enter(inode);

3718
	if (!ext4_can_truncate(inode))
3719 3720
		return;

3721
	ext4_clear_inode_flag(inode, EXT4_INODE_EOFBLOCKS);
3722

3723
	if (inode->i_size == 0 && !test_opt(inode->i_sb, NO_AUTO_DA_ALLOC))
3724
		ext4_set_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE);
3725

3726 3727 3728 3729 3730 3731 3732 3733
	if (ext4_has_inline_data(inode)) {
		int has_inline = 1;

		ext4_inline_data_truncate(inode, &has_inline);
		if (has_inline)
			return;
	}

3734 3735 3736 3737 3738 3739
	/* If we zero-out tail of the page, we have to create jinode for jbd2 */
	if (inode->i_size & (inode->i_sb->s_blocksize - 1)) {
		if (ext4_inode_attach_jinode(inode) < 0)
			return;
	}

T
Theodore Ts'o 已提交
3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750
	if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
		credits = ext4_writepage_trans_blocks(inode);
	else
		credits = ext4_blocks_for_truncate(inode);

	handle = ext4_journal_start(inode, EXT4_HT_TRUNCATE, credits);
	if (IS_ERR(handle)) {
		ext4_std_error(inode->i_sb, PTR_ERR(handle));
		return;
	}

3751 3752
	if (inode->i_size & (inode->i_sb->s_blocksize - 1))
		ext4_block_truncate_page(handle, mapping, inode->i_size);
T
Theodore Ts'o 已提交
3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769

	/*
	 * We add the inode to the orphan list, so that if this
	 * truncate spans multiple transactions, and we crash, we will
	 * resume the truncate when the filesystem recovers.  It also
	 * marks the inode dirty, to catch the new size.
	 *
	 * Implication: the file must always be in a sane, consistent
	 * truncatable state while each transaction commits.
	 */
	if (ext4_orphan_add(handle, inode))
		goto out_stop;

	down_write(&EXT4_I(inode)->i_data_sem);

	ext4_discard_preallocations(inode);

3770
	if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
T
Theodore Ts'o 已提交
3771
		ext4_ext_truncate(handle, inode);
3772
	else
T
Theodore Ts'o 已提交
3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793
		ext4_ind_truncate(handle, inode);

	up_write(&ei->i_data_sem);

	if (IS_SYNC(inode))
		ext4_handle_sync(handle);

out_stop:
	/*
	 * If this was a simple ftruncate() and the file will remain alive,
	 * then we need to clear up the orphan record which we created above.
	 * However, if this was a real unlink then we were called by
	 * ext4_delete_inode(), and we allow that function to clean up the
	 * orphan info for us.
	 */
	if (inode->i_nlink)
		ext4_orphan_del(handle, inode);

	inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
	ext4_mark_inode_dirty(handle, inode);
	ext4_journal_stop(handle);
3794

3795
	trace_ext4_truncate_exit(inode);
3796 3797 3798
}

/*
3799
 * ext4_get_inode_loc returns with an extra refcount against the inode's
3800 3801 3802 3803
 * underlying buffer_head on success. If 'in_mem' is true, we have all
 * data in memory that is needed to recreate the on-disk version of this
 * inode.
 */
3804 3805
static int __ext4_get_inode_loc(struct inode *inode,
				struct ext4_iloc *iloc, int in_mem)
3806
{
3807 3808 3809 3810 3811 3812
	struct ext4_group_desc	*gdp;
	struct buffer_head	*bh;
	struct super_block	*sb = inode->i_sb;
	ext4_fsblk_t		block;
	int			inodes_per_block, inode_offset;

A
Aneesh Kumar K.V 已提交
3813
	iloc->bh = NULL;
3814 3815
	if (!ext4_valid_inum(sb, inode->i_ino))
		return -EIO;
3816

3817 3818 3819
	iloc->block_group = (inode->i_ino - 1) / EXT4_INODES_PER_GROUP(sb);
	gdp = ext4_get_group_desc(sb, iloc->block_group, NULL);
	if (!gdp)
3820 3821
		return -EIO;

3822 3823 3824
	/*
	 * Figure out the offset within the block group inode table
	 */
3825
	inodes_per_block = EXT4_SB(sb)->s_inodes_per_block;
3826 3827 3828 3829 3830 3831
	inode_offset = ((inode->i_ino - 1) %
			EXT4_INODES_PER_GROUP(sb));
	block = ext4_inode_table(sb, gdp) + (inode_offset / inodes_per_block);
	iloc->offset = (inode_offset % inodes_per_block) * EXT4_INODE_SIZE(sb);

	bh = sb_getblk(sb, block);
3832
	if (unlikely(!bh))
3833
		return -ENOMEM;
3834 3835
	if (!buffer_uptodate(bh)) {
		lock_buffer(bh);
3836 3837 3838 3839 3840 3841 3842 3843 3844 3845

		/*
		 * If the buffer has the write error flag, we have failed
		 * to write out another inode in the same block.  In this
		 * case, we don't have to read the block because we may
		 * read the old inode data successfully.
		 */
		if (buffer_write_io_error(bh) && !buffer_uptodate(bh))
			set_buffer_uptodate(bh);

3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858
		if (buffer_uptodate(bh)) {
			/* someone brought it uptodate while we waited */
			unlock_buffer(bh);
			goto has_buffer;
		}

		/*
		 * If we have all information of the inode in memory and this
		 * is the only valid inode in the block, we need not read the
		 * block.
		 */
		if (in_mem) {
			struct buffer_head *bitmap_bh;
3859
			int i, start;
3860

3861
			start = inode_offset & ~(inodes_per_block - 1);
3862

3863 3864
			/* Is the inode bitmap in cache? */
			bitmap_bh = sb_getblk(sb, ext4_inode_bitmap(sb, gdp));
3865
			if (unlikely(!bitmap_bh))
3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876
				goto make_io;

			/*
			 * If the inode bitmap isn't in cache then the
			 * optimisation may end up performing two reads instead
			 * of one, so skip it.
			 */
			if (!buffer_uptodate(bitmap_bh)) {
				brelse(bitmap_bh);
				goto make_io;
			}
3877
			for (i = start; i < start + inodes_per_block; i++) {
3878 3879
				if (i == inode_offset)
					continue;
3880
				if (ext4_test_bit(i, bitmap_bh->b_data))
3881 3882 3883
					break;
			}
			brelse(bitmap_bh);
3884
			if (i == start + inodes_per_block) {
3885 3886 3887 3888 3889 3890 3891 3892 3893
				/* all other inodes are free, so skip I/O */
				memset(bh->b_data, 0, bh->b_size);
				set_buffer_uptodate(bh);
				unlock_buffer(bh);
				goto has_buffer;
			}
		}

make_io:
3894 3895 3896 3897 3898 3899 3900
		/*
		 * If we need to do any I/O, try to pre-readahead extra
		 * blocks from the inode table.
		 */
		if (EXT4_SB(sb)->s_inode_readahead_blks) {
			ext4_fsblk_t b, end, table;
			unsigned num;
3901
			__u32 ra_blks = EXT4_SB(sb)->s_inode_readahead_blks;
3902 3903

			table = ext4_inode_table(sb, gdp);
T
Theodore Ts'o 已提交
3904
			/* s_inode_readahead_blks is always a power of 2 */
3905
			b = block & ~((ext4_fsblk_t) ra_blks - 1);
3906 3907
			if (table > b)
				b = table;
3908
			end = b + ra_blks;
3909
			num = EXT4_INODES_PER_GROUP(sb);
3910
			if (ext4_has_group_desc_csum(sb))
3911
				num -= ext4_itable_unused_count(sb, gdp);
3912 3913 3914 3915 3916 3917 3918
			table += num / inodes_per_block;
			if (end > table)
				end = table;
			while (b <= end)
				sb_breadahead(sb, b++);
		}

3919 3920 3921 3922 3923
		/*
		 * There are other valid inodes in the buffer, this inode
		 * has in-inode xattrs, or we don't have this inode in memory.
		 * Read the block from disk.
		 */
3924
		trace_ext4_load_inode(inode);
3925 3926
		get_bh(bh);
		bh->b_end_io = end_buffer_read_sync;
3927
		submit_bh(READ | REQ_META | REQ_PRIO, bh);
3928 3929
		wait_on_buffer(bh);
		if (!buffer_uptodate(bh)) {
3930 3931
			EXT4_ERROR_INODE_BLOCK(inode, block,
					       "unable to read itable block");
3932 3933 3934 3935 3936 3937 3938 3939 3940
			brelse(bh);
			return -EIO;
		}
	}
has_buffer:
	iloc->bh = bh;
	return 0;
}

3941
int ext4_get_inode_loc(struct inode *inode, struct ext4_iloc *iloc)
3942 3943
{
	/* We have all inode data except xattrs in memory here. */
3944
	return __ext4_get_inode_loc(inode, iloc,
3945
		!ext4_test_inode_state(inode, EXT4_STATE_XATTR));
3946 3947
}

3948
void ext4_set_inode_flags(struct inode *inode)
3949
{
3950
	unsigned int flags = EXT4_I(inode)->i_flags;
3951 3952

	inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
3953
	if (flags & EXT4_SYNC_FL)
3954
		inode->i_flags |= S_SYNC;
3955
	if (flags & EXT4_APPEND_FL)
3956
		inode->i_flags |= S_APPEND;
3957
	if (flags & EXT4_IMMUTABLE_FL)
3958
		inode->i_flags |= S_IMMUTABLE;
3959
	if (flags & EXT4_NOATIME_FL)
3960
		inode->i_flags |= S_NOATIME;
3961
	if (flags & EXT4_DIRSYNC_FL)
3962 3963 3964
		inode->i_flags |= S_DIRSYNC;
}

3965 3966 3967
/* Propagate flags from i_flags to EXT4_I(inode)->i_flags */
void ext4_get_inode_flags(struct ext4_inode_info *ei)
{
3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987
	unsigned int vfs_fl;
	unsigned long old_fl, new_fl;

	do {
		vfs_fl = ei->vfs_inode.i_flags;
		old_fl = ei->i_flags;
		new_fl = old_fl & ~(EXT4_SYNC_FL|EXT4_APPEND_FL|
				EXT4_IMMUTABLE_FL|EXT4_NOATIME_FL|
				EXT4_DIRSYNC_FL);
		if (vfs_fl & S_SYNC)
			new_fl |= EXT4_SYNC_FL;
		if (vfs_fl & S_APPEND)
			new_fl |= EXT4_APPEND_FL;
		if (vfs_fl & S_IMMUTABLE)
			new_fl |= EXT4_IMMUTABLE_FL;
		if (vfs_fl & S_NOATIME)
			new_fl |= EXT4_NOATIME_FL;
		if (vfs_fl & S_DIRSYNC)
			new_fl |= EXT4_DIRSYNC_FL;
	} while (cmpxchg(&ei->i_flags, old_fl, new_fl) != old_fl);
3988
}
3989

3990
static blkcnt_t ext4_inode_blocks(struct ext4_inode *raw_inode,
3991
				  struct ext4_inode_info *ei)
3992 3993
{
	blkcnt_t i_blocks ;
A
Aneesh Kumar K.V 已提交
3994 3995
	struct inode *inode = &(ei->vfs_inode);
	struct super_block *sb = inode->i_sb;
3996 3997 3998 3999 4000 4001

	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
				EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
		/* we are using combined 48 bit field */
		i_blocks = ((u64)le16_to_cpu(raw_inode->i_blocks_high)) << 32 |
					le32_to_cpu(raw_inode->i_blocks_lo);
4002
		if (ext4_test_inode_flag(inode, EXT4_INODE_HUGE_FILE)) {
A
Aneesh Kumar K.V 已提交
4003 4004 4005 4006 4007
			/* i_blocks represent file system block size */
			return i_blocks  << (inode->i_blkbits - 9);
		} else {
			return i_blocks;
		}
4008 4009 4010 4011
	} else {
		return le32_to_cpu(raw_inode->i_blocks_lo);
	}
}
4012

4013 4014 4015 4016 4017 4018
static inline void ext4_iget_extra_inode(struct inode *inode,
					 struct ext4_inode *raw_inode,
					 struct ext4_inode_info *ei)
{
	__le32 *magic = (void *)raw_inode +
			EXT4_GOOD_OLD_INODE_SIZE + ei->i_extra_isize;
4019
	if (*magic == cpu_to_le32(EXT4_XATTR_MAGIC)) {
4020
		ext4_set_inode_state(inode, EXT4_STATE_XATTR);
4021
		ext4_find_inline_data_nolock(inode);
4022 4023
	} else
		EXT4_I(inode)->i_inline_off = 0;
4024 4025
}

4026
struct inode *ext4_iget(struct super_block *sb, unsigned long ino)
4027
{
4028 4029
	struct ext4_iloc iloc;
	struct ext4_inode *raw_inode;
4030 4031
	struct ext4_inode_info *ei;
	struct inode *inode;
4032
	journal_t *journal = EXT4_SB(sb)->s_journal;
4033
	long ret;
4034
	int block;
4035 4036
	uid_t i_uid;
	gid_t i_gid;
4037

4038 4039 4040 4041 4042 4043 4044
	inode = iget_locked(sb, ino);
	if (!inode)
		return ERR_PTR(-ENOMEM);
	if (!(inode->i_state & I_NEW))
		return inode;

	ei = EXT4_I(inode);
4045
	iloc.bh = NULL;
4046

4047 4048
	ret = __ext4_get_inode_loc(inode, &iloc, 0);
	if (ret < 0)
4049
		goto bad_inode;
4050
	raw_inode = ext4_raw_inode(&iloc);
4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083

	if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE) {
		ei->i_extra_isize = le16_to_cpu(raw_inode->i_extra_isize);
		if (EXT4_GOOD_OLD_INODE_SIZE + ei->i_extra_isize >
		    EXT4_INODE_SIZE(inode->i_sb)) {
			EXT4_ERROR_INODE(inode, "bad extra_isize (%u != %u)",
				EXT4_GOOD_OLD_INODE_SIZE + ei->i_extra_isize,
				EXT4_INODE_SIZE(inode->i_sb));
			ret = -EIO;
			goto bad_inode;
		}
	} else
		ei->i_extra_isize = 0;

	/* Precompute checksum seed for inode metadata */
	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
			EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
		struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
		__u32 csum;
		__le32 inum = cpu_to_le32(inode->i_ino);
		__le32 gen = raw_inode->i_generation;
		csum = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&inum,
				   sizeof(inum));
		ei->i_csum_seed = ext4_chksum(sbi, csum, (__u8 *)&gen,
					      sizeof(gen));
	}

	if (!ext4_inode_csum_verify(inode, raw_inode, ei)) {
		EXT4_ERROR_INODE(inode, "checksum invalid");
		ret = -EIO;
		goto bad_inode;
	}

4084
	inode->i_mode = le16_to_cpu(raw_inode->i_mode);
4085 4086
	i_uid = (uid_t)le16_to_cpu(raw_inode->i_uid_low);
	i_gid = (gid_t)le16_to_cpu(raw_inode->i_gid_low);
4087
	if (!(test_opt(inode->i_sb, NO_UID32))) {
4088 4089
		i_uid |= le16_to_cpu(raw_inode->i_uid_high) << 16;
		i_gid |= le16_to_cpu(raw_inode->i_gid_high) << 16;
4090
	}
4091 4092
	i_uid_write(inode, i_uid);
	i_gid_write(inode, i_gid);
M
Miklos Szeredi 已提交
4093
	set_nlink(inode, le16_to_cpu(raw_inode->i_links_count));
4094

4095
	ext4_clear_state_flags(ei);	/* Only relevant on 32-bit archs */
4096
	ei->i_inline_off = 0;
4097 4098 4099 4100 4101 4102 4103 4104
	ei->i_dir_start_lookup = 0;
	ei->i_dtime = le32_to_cpu(raw_inode->i_dtime);
	/* We now have enough fields to check if the inode was active or not.
	 * This is needed because nfsd might try to access dead inodes
	 * the test is that same one that e2fsck uses
	 * NeilBrown 1999oct15
	 */
	if (inode->i_nlink == 0) {
4105 4106 4107
		if ((inode->i_mode == 0 ||
		     !(EXT4_SB(inode->i_sb)->s_mount_state & EXT4_ORPHAN_FS)) &&
		    ino != EXT4_BOOT_LOADER_INO) {
4108
			/* this inode is deleted */
4109
			ret = -ESTALE;
4110 4111 4112 4113 4114
			goto bad_inode;
		}
		/* The only unlinked inodes we let through here have
		 * valid i_mode and are being read by the orphan
		 * recovery code: that's fine, we're about to complete
4115 4116 4117
		 * the process of deleting those.
		 * OR it is the EXT4_BOOT_LOADER_INO which is
		 * not initialized on a new filesystem. */
4118 4119
	}
	ei->i_flags = le32_to_cpu(raw_inode->i_flags);
4120
	inode->i_blocks = ext4_inode_blocks(raw_inode, ei);
4121
	ei->i_file_acl = le32_to_cpu(raw_inode->i_file_acl_lo);
4122
	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT))
B
Badari Pulavarty 已提交
4123 4124
		ei->i_file_acl |=
			((__u64)le16_to_cpu(raw_inode->i_file_acl_high)) << 32;
4125
	inode->i_size = ext4_isize(raw_inode);
4126
	ei->i_disksize = inode->i_size;
4127 4128 4129
#ifdef CONFIG_QUOTA
	ei->i_reserved_quota = 0;
#endif
4130 4131
	inode->i_generation = le32_to_cpu(raw_inode->i_generation);
	ei->i_block_group = iloc.block_group;
4132
	ei->i_last_alloc_group = ~0;
4133 4134 4135 4136
	/*
	 * NOTE! The in-memory inode i_data array is in little-endian order
	 * even on big-endian machines: we do NOT byteswap the block numbers!
	 */
4137
	for (block = 0; block < EXT4_N_BLOCKS; block++)
4138 4139 4140
		ei->i_data[block] = raw_inode->i_block[block];
	INIT_LIST_HEAD(&ei->i_orphan);

4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151
	/*
	 * Set transaction id's of transactions that have to be committed
	 * to finish f[data]sync. We set them to currently running transaction
	 * as we cannot be sure that the inode or some of its metadata isn't
	 * part of the transaction - the inode could have been reclaimed and
	 * now it is reread from disk.
	 */
	if (journal) {
		transaction_t *transaction;
		tid_t tid;

4152
		read_lock(&journal->j_state_lock);
4153 4154 4155 4156 4157 4158 4159 4160
		if (journal->j_running_transaction)
			transaction = journal->j_running_transaction;
		else
			transaction = journal->j_committing_transaction;
		if (transaction)
			tid = transaction->t_tid;
		else
			tid = journal->j_commit_sequence;
4161
		read_unlock(&journal->j_state_lock);
4162 4163 4164 4165
		ei->i_sync_tid = tid;
		ei->i_datasync_tid = tid;
	}

4166
	if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE) {
4167 4168
		if (ei->i_extra_isize == 0) {
			/* The extra space is currently unused. Use it. */
4169 4170
			ei->i_extra_isize = sizeof(struct ext4_inode) -
					    EXT4_GOOD_OLD_INODE_SIZE;
4171
		} else {
4172
			ext4_iget_extra_inode(inode, raw_inode, ei);
4173
		}
4174
	}
4175

K
Kalpak Shah 已提交
4176 4177 4178 4179 4180
	EXT4_INODE_GET_XTIME(i_ctime, inode, raw_inode);
	EXT4_INODE_GET_XTIME(i_mtime, inode, raw_inode);
	EXT4_INODE_GET_XTIME(i_atime, inode, raw_inode);
	EXT4_EINODE_GET_XTIME(i_crtime, ei, raw_inode);

4181 4182 4183 4184 4185 4186 4187
	inode->i_version = le32_to_cpu(raw_inode->i_disk_version);
	if (EXT4_INODE_SIZE(inode->i_sb) > EXT4_GOOD_OLD_INODE_SIZE) {
		if (EXT4_FITS_IN_INODE(raw_inode, ei, i_version_hi))
			inode->i_version |=
			(__u64)(le32_to_cpu(raw_inode->i_version_hi)) << 32;
	}

4188
	ret = 0;
4189
	if (ei->i_file_acl &&
4190
	    !ext4_data_block_valid(EXT4_SB(sb), ei->i_file_acl, 1)) {
4191 4192
		EXT4_ERROR_INODE(inode, "bad extended attribute block %llu",
				 ei->i_file_acl);
4193 4194
		ret = -EIO;
		goto bad_inode;
4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207
	} else if (!ext4_has_inline_data(inode)) {
		if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
			if ((S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
			    (S_ISLNK(inode->i_mode) &&
			     !ext4_inode_is_fast_symlink(inode))))
				/* Validate extent which is part of inode */
				ret = ext4_ext_check_inode(inode);
		} else if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
			   (S_ISLNK(inode->i_mode) &&
			    !ext4_inode_is_fast_symlink(inode))) {
			/* Validate block references which are part of inode */
			ret = ext4_ind_check_inode(inode);
		}
4208
	}
4209
	if (ret)
4210
		goto bad_inode;
4211

4212
	if (S_ISREG(inode->i_mode)) {
4213 4214 4215
		inode->i_op = &ext4_file_inode_operations;
		inode->i_fop = &ext4_file_operations;
		ext4_set_aops(inode);
4216
	} else if (S_ISDIR(inode->i_mode)) {
4217 4218
		inode->i_op = &ext4_dir_inode_operations;
		inode->i_fop = &ext4_dir_operations;
4219
	} else if (S_ISLNK(inode->i_mode)) {
4220
		if (ext4_inode_is_fast_symlink(inode)) {
4221
			inode->i_op = &ext4_fast_symlink_inode_operations;
4222 4223 4224
			nd_terminate_link(ei->i_data, inode->i_size,
				sizeof(ei->i_data) - 1);
		} else {
4225 4226
			inode->i_op = &ext4_symlink_inode_operations;
			ext4_set_aops(inode);
4227
		}
4228 4229
	} else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
	      S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
4230
		inode->i_op = &ext4_special_inode_operations;
4231 4232 4233 4234 4235 4236
		if (raw_inode->i_block[0])
			init_special_inode(inode, inode->i_mode,
			   old_decode_dev(le32_to_cpu(raw_inode->i_block[0])));
		else
			init_special_inode(inode, inode->i_mode,
			   new_decode_dev(le32_to_cpu(raw_inode->i_block[1])));
4237 4238
	} else if (ino == EXT4_BOOT_LOADER_INO) {
		make_bad_inode(inode);
4239 4240
	} else {
		ret = -EIO;
4241
		EXT4_ERROR_INODE(inode, "bogus i_mode (%o)", inode->i_mode);
4242
		goto bad_inode;
4243
	}
4244
	brelse(iloc.bh);
4245
	ext4_set_inode_flags(inode);
4246 4247
	unlock_new_inode(inode);
	return inode;
4248 4249

bad_inode:
4250
	brelse(iloc.bh);
4251 4252
	iget_failed(inode);
	return ERR_PTR(ret);
4253 4254
}

4255 4256 4257 4258 4259 4260 4261 4262 4263 4264
static int ext4_inode_blocks_set(handle_t *handle,
				struct ext4_inode *raw_inode,
				struct ext4_inode_info *ei)
{
	struct inode *inode = &(ei->vfs_inode);
	u64 i_blocks = inode->i_blocks;
	struct super_block *sb = inode->i_sb;

	if (i_blocks <= ~0U) {
		/*
4265
		 * i_blocks can be represented in a 32 bit variable
4266 4267
		 * as multiple of 512 bytes
		 */
A
Aneesh Kumar K.V 已提交
4268
		raw_inode->i_blocks_lo   = cpu_to_le32(i_blocks);
4269
		raw_inode->i_blocks_high = 0;
4270
		ext4_clear_inode_flag(inode, EXT4_INODE_HUGE_FILE);
4271 4272 4273 4274 4275 4276
		return 0;
	}
	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE))
		return -EFBIG;

	if (i_blocks <= 0xffffffffffffULL) {
4277 4278 4279 4280
		/*
		 * i_blocks can be represented in a 48 bit variable
		 * as multiple of 512 bytes
		 */
A
Aneesh Kumar K.V 已提交
4281
		raw_inode->i_blocks_lo   = cpu_to_le32(i_blocks);
4282
		raw_inode->i_blocks_high = cpu_to_le16(i_blocks >> 32);
4283
		ext4_clear_inode_flag(inode, EXT4_INODE_HUGE_FILE);
4284
	} else {
4285
		ext4_set_inode_flag(inode, EXT4_INODE_HUGE_FILE);
A
Aneesh Kumar K.V 已提交
4286 4287 4288 4289
		/* i_block is stored in file system block size */
		i_blocks = i_blocks >> (inode->i_blkbits - 9);
		raw_inode->i_blocks_lo   = cpu_to_le32(i_blocks);
		raw_inode->i_blocks_high = cpu_to_le16(i_blocks >> 32);
4290
	}
4291
	return 0;
4292 4293
}

4294 4295 4296 4297 4298 4299 4300
/*
 * Post the struct inode info into an on-disk inode location in the
 * buffer-cache.  This gobbles the caller's reference to the
 * buffer_head in the inode location struct.
 *
 * The caller must have write access to iloc->bh.
 */
4301
static int ext4_do_update_inode(handle_t *handle,
4302
				struct inode *inode,
4303
				struct ext4_iloc *iloc)
4304
{
4305 4306
	struct ext4_inode *raw_inode = ext4_raw_inode(iloc);
	struct ext4_inode_info *ei = EXT4_I(inode);
4307 4308
	struct buffer_head *bh = iloc->bh;
	int err = 0, rc, block;
4309
	int need_datasync = 0;
4310 4311
	uid_t i_uid;
	gid_t i_gid;
4312 4313 4314

	/* For fields not not tracking in the in-memory inode,
	 * initialise them to zero for new inodes. */
4315
	if (ext4_test_inode_state(inode, EXT4_STATE_NEW))
4316
		memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
4317

4318
	ext4_get_inode_flags(ei);
4319
	raw_inode->i_mode = cpu_to_le16(inode->i_mode);
4320 4321
	i_uid = i_uid_read(inode);
	i_gid = i_gid_read(inode);
4322
	if (!(test_opt(inode->i_sb, NO_UID32))) {
4323 4324
		raw_inode->i_uid_low = cpu_to_le16(low_16_bits(i_uid));
		raw_inode->i_gid_low = cpu_to_le16(low_16_bits(i_gid));
4325 4326 4327 4328
/*
 * Fix up interoperability with old kernels. Otherwise, old inodes get
 * re-used with the upper 16 bits of the uid/gid intact
 */
4329
		if (!ei->i_dtime) {
4330
			raw_inode->i_uid_high =
4331
				cpu_to_le16(high_16_bits(i_uid));
4332
			raw_inode->i_gid_high =
4333
				cpu_to_le16(high_16_bits(i_gid));
4334 4335 4336 4337 4338
		} else {
			raw_inode->i_uid_high = 0;
			raw_inode->i_gid_high = 0;
		}
	} else {
4339 4340
		raw_inode->i_uid_low = cpu_to_le16(fs_high2lowuid(i_uid));
		raw_inode->i_gid_low = cpu_to_le16(fs_high2lowgid(i_gid));
4341 4342 4343 4344
		raw_inode->i_uid_high = 0;
		raw_inode->i_gid_high = 0;
	}
	raw_inode->i_links_count = cpu_to_le16(inode->i_nlink);
K
Kalpak Shah 已提交
4345 4346 4347 4348 4349 4350

	EXT4_INODE_SET_XTIME(i_ctime, inode, raw_inode);
	EXT4_INODE_SET_XTIME(i_mtime, inode, raw_inode);
	EXT4_INODE_SET_XTIME(i_atime, inode, raw_inode);
	EXT4_EINODE_SET_XTIME(i_crtime, ei, raw_inode);

4351 4352
	if (ext4_inode_blocks_set(handle, raw_inode, ei))
		goto out_brelse;
4353
	raw_inode->i_dtime = cpu_to_le32(ei->i_dtime);
4354
	raw_inode->i_flags = cpu_to_le32(ei->i_flags & 0xFFFFFFFF);
4355 4356
	if (EXT4_SB(inode->i_sb)->s_es->s_creator_os !=
	    cpu_to_le32(EXT4_OS_HURD))
B
Badari Pulavarty 已提交
4357 4358
		raw_inode->i_file_acl_high =
			cpu_to_le16(ei->i_file_acl >> 32);
4359
	raw_inode->i_file_acl_lo = cpu_to_le32(ei->i_file_acl);
4360 4361 4362 4363
	if (ei->i_disksize != ext4_isize(raw_inode)) {
		ext4_isize_set(raw_inode, ei->i_disksize);
		need_datasync = 1;
	}
4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378
	if (ei->i_disksize > 0x7fffffffULL) {
		struct super_block *sb = inode->i_sb;
		if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
				EXT4_FEATURE_RO_COMPAT_LARGE_FILE) ||
				EXT4_SB(sb)->s_es->s_rev_level ==
				cpu_to_le32(EXT4_GOOD_OLD_REV)) {
			/* If this is the first large file
			 * created, add a flag to the superblock.
			 */
			err = ext4_journal_get_write_access(handle,
					EXT4_SB(sb)->s_sbh);
			if (err)
				goto out_brelse;
			ext4_update_dynamic_rev(sb);
			EXT4_SET_RO_COMPAT_FEATURE(sb,
4379
					EXT4_FEATURE_RO_COMPAT_LARGE_FILE);
4380
			ext4_handle_sync(handle);
4381
			err = ext4_handle_dirty_super(handle, sb);
4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395
		}
	}
	raw_inode->i_generation = cpu_to_le32(inode->i_generation);
	if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
		if (old_valid_dev(inode->i_rdev)) {
			raw_inode->i_block[0] =
				cpu_to_le32(old_encode_dev(inode->i_rdev));
			raw_inode->i_block[1] = 0;
		} else {
			raw_inode->i_block[0] = 0;
			raw_inode->i_block[1] =
				cpu_to_le32(new_encode_dev(inode->i_rdev));
			raw_inode->i_block[2] = 0;
		}
4396
	} else if (!ext4_has_inline_data(inode)) {
4397 4398
		for (block = 0; block < EXT4_N_BLOCKS; block++)
			raw_inode->i_block[block] = ei->i_data[block];
4399
	}
4400

4401 4402 4403 4404 4405
	raw_inode->i_disk_version = cpu_to_le32(inode->i_version);
	if (ei->i_extra_isize) {
		if (EXT4_FITS_IN_INODE(raw_inode, ei, i_version_hi))
			raw_inode->i_version_hi =
			cpu_to_le32(inode->i_version >> 32);
4406
		raw_inode->i_extra_isize = cpu_to_le16(ei->i_extra_isize);
4407 4408
	}

4409 4410
	ext4_inode_csum_set(inode, raw_inode, ei);

4411
	BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
4412
	rc = ext4_handle_dirty_metadata(handle, NULL, bh);
4413 4414
	if (!err)
		err = rc;
4415
	ext4_clear_inode_state(inode, EXT4_STATE_NEW);
4416

4417
	ext4_update_inode_fsync_trans(handle, inode, need_datasync);
4418
out_brelse:
4419
	brelse(bh);
4420
	ext4_std_error(inode->i_sb, err);
4421 4422 4423 4424
	return err;
}

/*
4425
 * ext4_write_inode()
4426 4427 4428 4429 4430
 *
 * We are called from a few places:
 *
 * - Within generic_file_write() for O_SYNC files.
 *   Here, there will be no transaction running. We wait for any running
4431
 *   transaction to commit.
4432 4433 4434 4435 4436 4437 4438 4439 4440 4441
 *
 * - Within sys_sync(), kupdate and such.
 *   We wait on commit, if tol to.
 *
 * - Within prune_icache() (PF_MEMALLOC == true)
 *   Here we simply return.  We can't afford to block kswapd on the
 *   journal commit.
 *
 * In all cases it is actually safe for us to return without doing anything,
 * because the inode has been copied into a raw inode buffer in
4442
 * ext4_mark_inode_dirty().  This is a correctness thing for O_SYNC and for
4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458
 * knfsd.
 *
 * Note that we are absolutely dependent upon all inode dirtiers doing the
 * right thing: they *must* call mark_inode_dirty() after dirtying info in
 * which we are interested.
 *
 * It would be a bug for them to not do this.  The code:
 *
 *	mark_inode_dirty(inode)
 *	stuff();
 *	inode->i_size = expr;
 *
 * is in error because a kswapd-driven write_inode() could occur while
 * `stuff()' is running, and the new i_size will be lost.  Plus the inode
 * will no longer be on the superblock's dirty inode list.
 */
4459
int ext4_write_inode(struct inode *inode, struct writeback_control *wbc)
4460
{
4461 4462
	int err;

4463 4464 4465
	if (current->flags & PF_MEMALLOC)
		return 0;

4466 4467 4468 4469 4470 4471
	if (EXT4_SB(inode->i_sb)->s_journal) {
		if (ext4_journal_current_handle()) {
			jbd_debug(1, "called recursively, non-PF_MEMALLOC!\n");
			dump_stack();
			return -EIO;
		}
4472

4473
		if (wbc->sync_mode != WB_SYNC_ALL)
4474 4475 4476 4477 4478
			return 0;

		err = ext4_force_commit(inode->i_sb);
	} else {
		struct ext4_iloc iloc;
4479

4480
		err = __ext4_get_inode_loc(inode, &iloc, 0);
4481 4482
		if (err)
			return err;
4483
		if (wbc->sync_mode == WB_SYNC_ALL)
4484 4485
			sync_dirty_buffer(iloc.bh);
		if (buffer_req(iloc.bh) && !buffer_uptodate(iloc.bh)) {
4486 4487
			EXT4_ERROR_INODE_BLOCK(inode, iloc.bh->b_blocknr,
					 "IO error syncing inode");
4488 4489
			err = -EIO;
		}
4490
		brelse(iloc.bh);
4491 4492
	}
	return err;
4493 4494
}

4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520
/*
 * In data=journal mode ext4_journalled_invalidatepage() may fail to invalidate
 * buffers that are attached to a page stradding i_size and are undergoing
 * commit. In that case we have to wait for commit to finish and try again.
 */
static void ext4_wait_for_tail_page_commit(struct inode *inode)
{
	struct page *page;
	unsigned offset;
	journal_t *journal = EXT4_SB(inode->i_sb)->s_journal;
	tid_t commit_tid = 0;
	int ret;

	offset = inode->i_size & (PAGE_CACHE_SIZE - 1);
	/*
	 * All buffers in the last page remain valid? Then there's nothing to
	 * do. We do the check mainly to optimize the common PAGE_CACHE_SIZE ==
	 * blocksize case
	 */
	if (offset > PAGE_CACHE_SIZE - (1 << inode->i_blkbits))
		return;
	while (1) {
		page = find_lock_page(inode->i_mapping,
				      inode->i_size >> PAGE_CACHE_SHIFT);
		if (!page)
			return;
4521 4522
		ret = __ext4_journalled_invalidatepage(page, offset,
						PAGE_CACHE_SIZE - offset);
4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536
		unlock_page(page);
		page_cache_release(page);
		if (ret != -EBUSY)
			return;
		commit_tid = 0;
		read_lock(&journal->j_state_lock);
		if (journal->j_committing_transaction)
			commit_tid = journal->j_committing_transaction->t_tid;
		read_unlock(&journal->j_state_lock);
		if (commit_tid)
			jbd2_log_wait_commit(journal, commit_tid);
	}
}

4537
/*
4538
 * ext4_setattr()
4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551
 *
 * Called from notify_change.
 *
 * We want to trap VFS attempts to truncate the file as soon as
 * possible.  In particular, we want to make sure that when the VFS
 * shrinks i_size, we put the inode on the orphan list and modify
 * i_disksize immediately, so that during the subsequent flushing of
 * dirty pages and freeing of disk blocks, we can guarantee that any
 * commit will leave the blocks being flushed in an unused state on
 * disk.  (On recovery, the inode will get truncated and the blocks will
 * be freed, so we have a strong guarantee that no future commit will
 * leave these blocks visible to the user.)
 *
4552 4553 4554 4555 4556 4557 4558 4559
 * Another thing we have to assure is that if we are in ordered mode
 * and inode is still attached to the committing transaction, we must
 * we start writeout of all the dirty pages which are being truncated.
 * This way we are sure that all the data written in the previous
 * transaction are already on disk (truncate waits for pages under
 * writeback).
 *
 * Called with inode->i_mutex down.
4560
 */
4561
int ext4_setattr(struct dentry *dentry, struct iattr *attr)
4562 4563 4564
{
	struct inode *inode = dentry->d_inode;
	int error, rc = 0;
4565
	int orphan = 0;
4566 4567 4568 4569 4570 4571
	const unsigned int ia_valid = attr->ia_valid;

	error = inode_change_ok(inode, attr);
	if (error)
		return error;

4572
	if (is_quota_modification(inode, attr))
4573
		dquot_initialize(inode);
4574 4575
	if ((ia_valid & ATTR_UID && !uid_eq(attr->ia_uid, inode->i_uid)) ||
	    (ia_valid & ATTR_GID && !gid_eq(attr->ia_gid, inode->i_gid))) {
4576 4577 4578 4579
		handle_t *handle;

		/* (user+group)*(old+new) structure, inode write (sb,
		 * inode block, ? - but truncate inode update has it) */
4580 4581 4582
		handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
			(EXT4_MAXQUOTAS_INIT_BLOCKS(inode->i_sb) +
			 EXT4_MAXQUOTAS_DEL_BLOCKS(inode->i_sb)) + 3);
4583 4584 4585 4586
		if (IS_ERR(handle)) {
			error = PTR_ERR(handle);
			goto err_out;
		}
4587
		error = dquot_transfer(inode, attr);
4588
		if (error) {
4589
			ext4_journal_stop(handle);
4590 4591 4592 4593 4594 4595 4596 4597
			return error;
		}
		/* Update corresponding info in inode so that everything is in
		 * one transaction */
		if (attr->ia_valid & ATTR_UID)
			inode->i_uid = attr->ia_uid;
		if (attr->ia_valid & ATTR_GID)
			inode->i_gid = attr->ia_gid;
4598 4599
		error = ext4_mark_inode_dirty(handle, inode);
		ext4_journal_stop(handle);
4600 4601
	}

4602 4603 4604
	if (attr->ia_valid & ATTR_SIZE && attr->ia_size != inode->i_size) {
		handle_t *handle;
		loff_t oldsize = inode->i_size;
4605

4606
		if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) {
4607 4608
			struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);

4609 4610
			if (attr->ia_size > sbi->s_bitmap_maxbytes)
				return -EFBIG;
4611
		}
4612 4613 4614 4615
		if (S_ISREG(inode->i_mode) &&
		    (attr->ia_size < inode->i_size)) {
			if (ext4_should_order_data(inode)) {
				error = ext4_begin_ordered_truncate(inode,
4616
							    attr->ia_size);
4617
				if (error)
4618
					goto err_out;
4619 4620 4621 4622 4623 4624 4625 4626 4627 4628
			}
			handle = ext4_journal_start(inode, EXT4_HT_INODE, 3);
			if (IS_ERR(handle)) {
				error = PTR_ERR(handle);
				goto err_out;
			}
			if (ext4_handle_valid(handle)) {
				error = ext4_orphan_add(handle, inode);
				orphan = 1;
			}
4629
			down_write(&EXT4_I(inode)->i_data_sem);
4630 4631 4632 4633
			EXT4_I(inode)->i_disksize = attr->ia_size;
			rc = ext4_mark_inode_dirty(handle, inode);
			if (!error)
				error = rc;
4634 4635 4636 4637 4638 4639 4640 4641
			/*
			 * We have to update i_size under i_data_sem together
			 * with i_disksize to avoid races with writeback code
			 * running ext4_wb_update_i_disksize().
			 */
			if (!error)
				i_size_write(inode, attr->ia_size);
			up_write(&EXT4_I(inode)->i_data_sem);
4642 4643 4644
			ext4_journal_stop(handle);
			if (error) {
				ext4_orphan_del(NULL, inode);
4645 4646
				goto err_out;
			}
4647 4648
		} else
			i_size_write(inode, attr->ia_size);
4649

4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661
		/*
		 * Blocks are going to be removed from the inode. Wait
		 * for dio in flight.  Temporarily disable
		 * dioread_nolock to prevent livelock.
		 */
		if (orphan) {
			if (!ext4_should_journal_data(inode)) {
				ext4_inode_block_unlocked_dio(inode);
				inode_dio_wait(inode);
				ext4_inode_resume_unlocked_dio(inode);
			} else
				ext4_wait_for_tail_page_commit(inode);
4662
		}
4663 4664 4665 4666 4667
		/*
		 * Truncate pagecache after we've waited for commit
		 * in data=journal mode to make pages freeable.
		 */
		truncate_pagecache(inode, oldsize, inode->i_size);
4668
	}
4669 4670 4671 4672 4673 4674
	/*
	 * We want to call ext4_truncate() even if attr->ia_size ==
	 * inode->i_size for cases like truncation of fallocated space
	 */
	if (attr->ia_valid & ATTR_SIZE)
		ext4_truncate(inode);
4675

C
Christoph Hellwig 已提交
4676 4677 4678 4679 4680 4681 4682 4683 4684
	if (!rc) {
		setattr_copy(inode, attr);
		mark_inode_dirty(inode);
	}

	/*
	 * If the call to ext4_truncate failed to get a transaction handle at
	 * all, we need to clean up the in-core orphan list manually.
	 */
4685
	if (orphan && inode->i_nlink)
4686
		ext4_orphan_del(NULL, inode);
4687 4688

	if (!rc && (ia_valid & ATTR_MODE))
4689
		rc = ext4_acl_chmod(inode);
4690 4691

err_out:
4692
	ext4_std_error(inode->i_sb, error);
4693 4694 4695 4696 4697
	if (!error)
		error = rc;
	return error;
}

4698 4699 4700 4701
int ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
		 struct kstat *stat)
{
	struct inode *inode;
4702
	unsigned long long delalloc_blocks;
4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716

	inode = dentry->d_inode;
	generic_fillattr(inode, stat);

	/*
	 * We can't update i_blocks if the block allocation is delayed
	 * otherwise in the case of system crash before the real block
	 * allocation is done, we will have i_blocks inconsistent with
	 * on-disk file blocks.
	 * We always keep i_blocks updated together with real
	 * allocation. But to not confuse with user, stat
	 * will return the blocks that include the delayed allocation
	 * blocks for this file.
	 */
4717 4718
	delalloc_blocks = EXT4_C2B(EXT4_SB(inode->i_sb),
				EXT4_I(inode)->i_reserved_data_blocks);
4719

4720
	stat->blocks += delalloc_blocks << (inode->i_sb->s_blocksize_bits-9);
4721 4722
	return 0;
}
4723

4724 4725
static int ext4_index_trans_blocks(struct inode *inode, int lblocks,
				   int pextents)
4726
{
4727
	if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
4728 4729
		return ext4_ind_trans_blocks(inode, lblocks);
	return ext4_ext_index_trans_blocks(inode, pextents);
4730
}
4731

4732
/*
4733 4734 4735
 * Account for index blocks, block groups bitmaps and block group
 * descriptor blocks if modify datablocks and index blocks
 * worse case, the indexs blocks spread over different block groups
4736
 *
4737
 * If datablocks are discontiguous, they are possible to spread over
4738
 * different block groups too. If they are contiguous, with flexbg,
4739
 * they could still across block group boundary.
4740
 *
4741 4742
 * Also account for superblock, inode, quota and xattr blocks
 */
4743 4744
static int ext4_meta_trans_blocks(struct inode *inode, int lblocks,
				  int pextents)
4745
{
4746 4747
	ext4_group_t groups, ngroups = ext4_get_groups_count(inode->i_sb);
	int gdpblocks;
4748 4749 4750 4751
	int idxblocks;
	int ret = 0;

	/*
4752 4753
	 * How many index blocks need to touch to map @lblocks logical blocks
	 * to @pextents physical extents?
4754
	 */
4755
	idxblocks = ext4_index_trans_blocks(inode, lblocks, pextents);
4756 4757 4758 4759 4760 4761 4762

	ret = idxblocks;

	/*
	 * Now let's see how many group bitmaps and group descriptors need
	 * to account
	 */
4763
	groups = idxblocks + pextents;
4764
	gdpblocks = groups;
4765 4766
	if (groups > ngroups)
		groups = ngroups;
4767 4768 4769 4770 4771 4772 4773 4774 4775 4776 4777 4778 4779
	if (groups > EXT4_SB(inode->i_sb)->s_gdb_count)
		gdpblocks = EXT4_SB(inode->i_sb)->s_gdb_count;

	/* bitmaps and block group descriptor blocks */
	ret += groups + gdpblocks;

	/* Blocks for super block, inode, quota and xattr blocks */
	ret += EXT4_META_TRANS_BLOCKS(inode->i_sb);

	return ret;
}

/*
L
Lucas De Marchi 已提交
4780
 * Calculate the total number of credits to reserve to fit
4781 4782
 * the modification of a single pages into a single transaction,
 * which may include multiple chunks of block allocations.
4783
 *
4784
 * This could be called via ext4_write_begin()
4785
 *
4786
 * We need to consider the worse case, when
4787
 * one new block per extent.
4788
 */
A
Alex Tomas 已提交
4789
int ext4_writepage_trans_blocks(struct inode *inode)
4790
{
4791
	int bpp = ext4_journal_blocks_per_page(inode);
4792 4793
	int ret;

4794
	ret = ext4_meta_trans_blocks(inode, bpp, bpp);
A
Alex Tomas 已提交
4795

4796
	/* Account for data blocks for journalled mode */
4797
	if (ext4_should_journal_data(inode))
4798
		ret += bpp;
4799 4800
	return ret;
}
4801 4802 4803 4804 4805

/*
 * Calculate the journal credits for a chunk of data modification.
 *
 * This is called from DIO, fallocate or whoever calling
4806
 * ext4_map_blocks() to map/allocate a chunk of contiguous disk blocks.
4807 4808 4809 4810 4811 4812 4813 4814 4815
 *
 * journal buffers for data blocks are not included here, as DIO
 * and fallocate do no need to journal data buffers.
 */
int ext4_chunk_trans_blocks(struct inode *inode, int nrblocks)
{
	return ext4_meta_trans_blocks(inode, nrblocks, 1);
}

4816
/*
4817
 * The caller must have previously called ext4_reserve_inode_write().
4818 4819
 * Give this, we know that the caller already has write access to iloc->bh.
 */
4820
int ext4_mark_iloc_dirty(handle_t *handle,
4821
			 struct inode *inode, struct ext4_iloc *iloc)
4822 4823 4824
{
	int err = 0;

4825
	if (IS_I_VERSION(inode))
4826 4827
		inode_inc_iversion(inode);

4828 4829 4830
	/* the do_update_inode consumes one bh->b_count */
	get_bh(iloc->bh);

4831
	/* ext4_do_update_inode() does jbd2_journal_dirty_metadata */
4832
	err = ext4_do_update_inode(handle, inode, iloc);
4833 4834 4835 4836 4837 4838 4839 4840 4841 4842
	put_bh(iloc->bh);
	return err;
}

/*
 * On success, We end up with an outstanding reference count against
 * iloc->bh.  This _must_ be cleaned up later.
 */

int
4843 4844
ext4_reserve_inode_write(handle_t *handle, struct inode *inode,
			 struct ext4_iloc *iloc)
4845
{
4846 4847 4848 4849 4850 4851 4852 4853 4854
	int err;

	err = ext4_get_inode_loc(inode, iloc);
	if (!err) {
		BUFFER_TRACE(iloc->bh, "get_write_access");
		err = ext4_journal_get_write_access(handle, iloc->bh);
		if (err) {
			brelse(iloc->bh);
			iloc->bh = NULL;
4855 4856
		}
	}
4857
	ext4_std_error(inode->i_sb, err);
4858 4859 4860
	return err;
}

4861 4862 4863 4864
/*
 * Expand an inode by new_extra_isize bytes.
 * Returns 0 on success or negative error number on failure.
 */
A
Aneesh Kumar K.V 已提交
4865 4866 4867 4868
static int ext4_expand_extra_isize(struct inode *inode,
				   unsigned int new_extra_isize,
				   struct ext4_iloc iloc,
				   handle_t *handle)
4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879 4880
{
	struct ext4_inode *raw_inode;
	struct ext4_xattr_ibody_header *header;

	if (EXT4_I(inode)->i_extra_isize >= new_extra_isize)
		return 0;

	raw_inode = ext4_raw_inode(&iloc);

	header = IHDR(inode, raw_inode);

	/* No extended attributes present */
4881 4882
	if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR) ||
	    header->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC)) {
4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893
		memset((void *)raw_inode + EXT4_GOOD_OLD_INODE_SIZE, 0,
			new_extra_isize);
		EXT4_I(inode)->i_extra_isize = new_extra_isize;
		return 0;
	}

	/* try to expand with EAs present */
	return ext4_expand_extra_isize_ea(inode, new_extra_isize,
					  raw_inode, handle);
}

4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906
/*
 * What we do here is to mark the in-core inode as clean with respect to inode
 * dirtiness (it may still be data-dirty).
 * This means that the in-core inode may be reaped by prune_icache
 * without having to perform any I/O.  This is a very good thing,
 * because *any* task may call prune_icache - even ones which
 * have a transaction open against a different journal.
 *
 * Is this cheating?  Not really.  Sure, we haven't written the
 * inode out, but prune_icache isn't a user-visible syncing function.
 * Whenever the user wants stuff synced (sys_sync, sys_msync, sys_fsync)
 * we start and wait on commits.
 */
4907
int ext4_mark_inode_dirty(handle_t *handle, struct inode *inode)
4908
{
4909
	struct ext4_iloc iloc;
4910 4911 4912
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
	static unsigned int mnt_count;
	int err, ret;
4913 4914

	might_sleep();
4915
	trace_ext4_mark_inode_dirty(inode, _RET_IP_);
4916
	err = ext4_reserve_inode_write(handle, inode, &iloc);
4917 4918
	if (ext4_handle_valid(handle) &&
	    EXT4_I(inode)->i_extra_isize < sbi->s_want_extra_isize &&
4919
	    !ext4_test_inode_state(inode, EXT4_STATE_NO_EXPAND)) {
4920 4921 4922 4923 4924 4925 4926 4927 4928 4929 4930 4931 4932
		/*
		 * We need extra buffer credits since we may write into EA block
		 * with this same handle. If journal_extend fails, then it will
		 * only result in a minor loss of functionality for that inode.
		 * If this is felt to be critical, then e2fsck should be run to
		 * force a large enough s_min_extra_isize.
		 */
		if ((jbd2_journal_extend(handle,
			     EXT4_DATA_TRANS_BLOCKS(inode->i_sb))) == 0) {
			ret = ext4_expand_extra_isize(inode,
						      sbi->s_want_extra_isize,
						      iloc, handle);
			if (ret) {
4933 4934
				ext4_set_inode_state(inode,
						     EXT4_STATE_NO_EXPAND);
A
Aneesh Kumar K.V 已提交
4935 4936
				if (mnt_count !=
					le16_to_cpu(sbi->s_es->s_mnt_count)) {
4937
					ext4_warning(inode->i_sb,
4938 4939 4940
					"Unable to expand inode %lu. Delete"
					" some EAs or run e2fsck.",
					inode->i_ino);
A
Aneesh Kumar K.V 已提交
4941 4942
					mnt_count =
					  le16_to_cpu(sbi->s_es->s_mnt_count);
4943 4944 4945 4946
				}
			}
		}
	}
4947
	if (!err)
4948
		err = ext4_mark_iloc_dirty(handle, inode, &iloc);
4949 4950 4951 4952
	return err;
}

/*
4953
 * ext4_dirty_inode() is called from __mark_inode_dirty()
4954 4955 4956 4957 4958
 *
 * We're really interested in the case where a file is being extended.
 * i_size has been changed by generic_commit_write() and we thus need
 * to include the updated inode in the current transaction.
 *
4959
 * Also, dquot_alloc_block() will always dirty the inode when blocks
4960 4961 4962 4963 4964 4965
 * are allocated to the file.
 *
 * If the inode is marked synchronous, we don't honour that here - doing
 * so would cause a commit on atime updates, which we don't bother doing.
 * We handle synchronous inodes at the highest possible level.
 */
4966
void ext4_dirty_inode(struct inode *inode, int flags)
4967 4968 4969
{
	handle_t *handle;

4970
	handle = ext4_journal_start(inode, EXT4_HT_INODE, 2);
4971 4972
	if (IS_ERR(handle))
		goto out;
4973 4974 4975

	ext4_mark_inode_dirty(handle, inode);

4976
	ext4_journal_stop(handle);
4977 4978 4979 4980 4981 4982 4983 4984
out:
	return;
}

#if 0
/*
 * Bind an inode's backing buffer_head into this transaction, to prevent
 * it from being flushed to disk early.  Unlike
4985
 * ext4_reserve_inode_write, this leaves behind no bh reference and
4986 4987 4988
 * returns no iloc structure, so the caller needs to repeat the iloc
 * lookup to mark the inode dirty later.
 */
4989
static int ext4_pin_inode(handle_t *handle, struct inode *inode)
4990
{
4991
	struct ext4_iloc iloc;
4992 4993 4994

	int err = 0;
	if (handle) {
4995
		err = ext4_get_inode_loc(inode, &iloc);
4996 4997
		if (!err) {
			BUFFER_TRACE(iloc.bh, "get_write_access");
4998
			err = jbd2_journal_get_write_access(handle, iloc.bh);
4999
			if (!err)
5000
				err = ext4_handle_dirty_metadata(handle,
5001
								 NULL,
5002
								 iloc.bh);
5003 5004 5005
			brelse(iloc.bh);
		}
	}
5006
	ext4_std_error(inode->i_sb, err);
5007 5008 5009 5010
	return err;
}
#endif

5011
int ext4_change_inode_journal_flag(struct inode *inode, int val)
5012 5013 5014 5015 5016 5017 5018 5019 5020 5021 5022 5023 5024 5025 5026
{
	journal_t *journal;
	handle_t *handle;
	int err;

	/*
	 * We have to be very careful here: changing a data block's
	 * journaling status dynamically is dangerous.  If we write a
	 * data block to the journal, change the status and then delete
	 * that block, we risk forgetting to revoke the old log record
	 * from the journal and so a subsequent replay can corrupt data.
	 * So, first we make sure that the journal is empty and that
	 * nobody is changing anything.
	 */

5027
	journal = EXT4_JOURNAL(inode);
5028 5029
	if (!journal)
		return 0;
5030
	if (is_journal_aborted(journal))
5031
		return -EROFS;
5032 5033 5034 5035 5036 5037 5038 5039 5040 5041 5042
	/* We have to allocate physical blocks for delalloc blocks
	 * before flushing journal. otherwise delalloc blocks can not
	 * be allocated any more. even more truncate on delalloc blocks
	 * could trigger BUG by flushing delalloc blocks in journal.
	 * There is no delalloc block in non-journal data mode.
	 */
	if (val && test_opt(inode->i_sb, DELALLOC)) {
		err = ext4_alloc_da_blocks(inode);
		if (err < 0)
			return err;
	}
5043

5044 5045 5046 5047
	/* Wait for all existing dio workers */
	ext4_inode_block_unlocked_dio(inode);
	inode_dio_wait(inode);

5048
	jbd2_journal_lock_updates(journal);
5049 5050 5051 5052 5053 5054 5055 5056 5057 5058

	/*
	 * OK, there are no updates running now, and all cached data is
	 * synced to disk.  We are now in a completely consistent state
	 * which doesn't have anything in the journal, and we know that
	 * no filesystem updates are running, so it is safe to modify
	 * the inode's in-core data-journaling state flag now.
	 */

	if (val)
5059
		ext4_set_inode_flag(inode, EXT4_INODE_JOURNAL_DATA);
5060 5061
	else {
		jbd2_journal_flush(journal);
5062
		ext4_clear_inode_flag(inode, EXT4_INODE_JOURNAL_DATA);
5063
	}
5064
	ext4_set_aops(inode);
5065

5066
	jbd2_journal_unlock_updates(journal);
5067
	ext4_inode_resume_unlocked_dio(inode);
5068 5069 5070

	/* Finally we can mark the inode as dirty. */

5071
	handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
5072 5073 5074
	if (IS_ERR(handle))
		return PTR_ERR(handle);

5075
	err = ext4_mark_inode_dirty(handle, inode);
5076
	ext4_handle_sync(handle);
5077 5078
	ext4_journal_stop(handle);
	ext4_std_error(inode->i_sb, err);
5079 5080 5081

	return err;
}
5082 5083 5084 5085 5086 5087

static int ext4_bh_unmapped(handle_t *handle, struct buffer_head *bh)
{
	return !buffer_mapped(bh);
}

5088
int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
5089
{
5090
	struct page *page = vmf->page;
5091 5092
	loff_t size;
	unsigned long len;
5093
	int ret;
5094
	struct file *file = vma->vm_file;
A
Al Viro 已提交
5095
	struct inode *inode = file_inode(file);
5096
	struct address_space *mapping = inode->i_mapping;
5097 5098 5099
	handle_t *handle;
	get_block_t *get_block;
	int retries = 0;
5100

5101
	sb_start_pagefault(inode->i_sb);
5102
	file_update_time(vma->vm_file);
5103 5104 5105 5106 5107 5108 5109 5110 5111 5112
	/* Delalloc case is easy... */
	if (test_opt(inode->i_sb, DELALLOC) &&
	    !ext4_should_journal_data(inode) &&
	    !ext4_nonda_switch(inode->i_sb)) {
		do {
			ret = __block_page_mkwrite(vma, vmf,
						   ext4_da_get_block_prep);
		} while (ret == -ENOSPC &&
		       ext4_should_retry_alloc(inode->i_sb, &retries));
		goto out_ret;
5113
	}
5114 5115

	lock_page(page);
5116 5117 5118 5119 5120 5121
	size = i_size_read(inode);
	/* Page got truncated from under us? */
	if (page->mapping != mapping || page_offset(page) > size) {
		unlock_page(page);
		ret = VM_FAULT_NOPAGE;
		goto out;
5122
	}
5123 5124 5125 5126 5127

	if (page->index == size >> PAGE_CACHE_SHIFT)
		len = size & ~PAGE_CACHE_MASK;
	else
		len = PAGE_CACHE_SIZE;
5128
	/*
5129 5130
	 * Return if we have all the buffers mapped. This avoids the need to do
	 * journal_start/journal_stop which can block and take a long time
5131
	 */
5132
	if (page_has_buffers(page)) {
5133 5134 5135
		if (!ext4_walk_page_buffers(NULL, page_buffers(page),
					    0, len, NULL,
					    ext4_bh_unmapped)) {
5136
			/* Wait so that we don't change page under IO */
5137
			wait_for_stable_page(page);
5138 5139
			ret = VM_FAULT_LOCKED;
			goto out;
5140
		}
5141
	}
5142
	unlock_page(page);
5143 5144 5145 5146 5147 5148
	/* OK, we need to fill the hole... */
	if (ext4_should_dioread_nolock(inode))
		get_block = ext4_get_block_write;
	else
		get_block = ext4_get_block;
retry_alloc:
5149 5150
	handle = ext4_journal_start(inode, EXT4_HT_WRITE_PAGE,
				    ext4_writepage_trans_blocks(inode));
5151
	if (IS_ERR(handle)) {
5152
		ret = VM_FAULT_SIGBUS;
5153 5154 5155 5156
		goto out;
	}
	ret = __block_page_mkwrite(vma, vmf, get_block);
	if (!ret && ext4_should_journal_data(inode)) {
5157
		if (ext4_walk_page_buffers(handle, page_buffers(page), 0,
5158 5159 5160
			  PAGE_CACHE_SIZE, NULL, do_journal_get_write_access)) {
			unlock_page(page);
			ret = VM_FAULT_SIGBUS;
5161
			ext4_journal_stop(handle);
5162 5163 5164 5165 5166 5167 5168 5169 5170 5171
			goto out;
		}
		ext4_set_inode_state(inode, EXT4_STATE_JDATA);
	}
	ext4_journal_stop(handle);
	if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
		goto retry_alloc;
out_ret:
	ret = block_page_mkwrite_return(ret);
out:
5172
	sb_end_pagefault(inode->i_sb);
5173 5174
	return ret;
}